Cander lets you send, receive and save greeting cards in a fun and meaningful way

Cander

cander

Star Wars Knights of the Old Republic II: The Sith Lords (Kotor 2) is one of the most popular RPGs of all time. The famous Star Wars game is the direct sequel to the original Kotor. After you reach a certain point in the story, you are given the option of exploring Kotor 2's galaxy however you want, with the opportunity to choose which planets you will traverse. Here is the best Kotor 2 planet order and the best way to travel the stars. The Best Kotor 2 Planet Order is: Na Shaddaa Ondoron/ Dxun Part 1 Dantooine Ondoron/ Dxun Part 2 Korriban Kotor 2 has a vast range of planets available to you to explore. After you make it through a specific part of the game, the world of Kotor 2 opens up, allowing you to travel to a vast range of planets in any order you want. Such planets include the Sith homeworld Korriban, Na Shaddaa, Onderon, Dxun, and Dantooine. It is essential to traverse these planets properly so that you can enjoy the story better and maximise your characters' abilities before you reach the end of the game. Kotor 2 Planet Order Guide 1.Na Shaddaa The first planet you visit should be Na Shaddaa. Na Shaddaa is a moon that orbits the Hutt homeworld of Nal Hutta. The planet is called the smuggler's moon and is run by the infamous Hutt Cartel. Thus, the planet is full of corruption and illegal activities. Once on the planet, the game has your character defeat bounty hunters, save a Jedi master, and stop a criminal organisation. Thus, it provides an excellent introduction to the world of Kotor 2 without being super challenging, allowing you to level up for more complicated parts of the game. This planet should be the first place you visit on your journey due to your activities on the planet setting up nicely your future endeavours in-game. 2.Ondoron/ Dxun Part 1 After leaving Na Shaddaa, you should make your way to Dxun, a jungle moon located near Ondoron. Dxun is a moon inhabited by dangerous wildlife; after landing on the moon, your character is quickly mobbed by creatures such as the Zakkeg. After making your way through the moon and meeting Mandalore Canderous Ordo, the game has your character find a Jedi master in hiding on the planet of Onderon. After completing this mission, Canderous Ordo decides to join your team, aiding you in your adventure. Thus, traveling to these two planets first is the best way to travel, allowing you to not only level up but gain a powerful new ally in the fight against the Sith. Canderous Ordo is one of the best shooters in the game and will be vital in helping you to defeat the game's most complex challenges. 3.Dantooine To follow the best Kotor 2 planet order, you should next travel to Dantooine after recruiting Canderous on Dxun/Ondoron. Dantooine is a planet initially introduced in the first Kotor game. The world was home to a Jedi enclave that helped turn Darth Revan toward the light side of the force. However, it was destroyed by Darth Malek in subsequent years. Once on the planet, the game has you and your companions help defend against a Sith onslaught alongside Jedi Master Vrook Lamar. Dantooine is the next best place to visit on this list due to not only offering a moderate level of challenge for the final game boss but bridging the gap between the first and the second game in preparation for Kotor 2's final story reveals. 4.Ondoron/ Dxun Part 2 After saving Dantooine from the Sith, you should revisit the Ondoron/Dxun story arc. Once back on the two planets, the game has you defeat several Sith and stop a civil war on Onderon. Revisiting the two planets is the best place to set up several story arcs that will play out later in the story. The story arc also offers considerable challenge making it the ideal place to visit near the end of the game. 5. Korriban The final planet in our Kotor 2 planet order is Korriban; it should be the last place you visit in the game. Korriban was initially present in the first Kotor game and is known as the homeworld of the Sith. The game has you fight ancient Sith lords searching for a Jedi master, which quickly leads into an ambush by Darth Sion, one of the game's main villains. Leading Korriban to last is ideal due to the planet’s difficulty and the boss fight with Sion. Also, it perfectly sets up the game’s end sequence and story arc. Final Thoughts Overall, the Kotor 2 planet order above is the best way to traverse the stars and story. With the likely release of a Kotor remake shortly, inevitably, a Kotor 2 may also get remade, making this guide even more beneficial. For more Star Wars gaming content, . subscribe to our newsletter below Also, check out, 5 Best Minecraft Hacks and Cheats: How to Enable and Use Them 3 Best Kotor Builds Even Vader Would Approve of How to Get Anarchy in Destiny 2 (Destiny Exotic Guide)

Ultimate KOTOR 2 Planet Order: The Best Way to Traverse the Stars 

Ultimate KOTOR 2 Planet Order: la mejor manera de atravesar las estrellas

SAN FRANSCISCO, United States, May 24th, 2024/Chainwire/ Rollup-as-a-Service platform Caldera is launching Guardian Nodes, a production-ready node system that enables network users to verify rollup blocks for rewards.
HYCHAIN, the first team to leverage this system with Caldera, generated $8 million from their initial node sale. Caldera, the leading Rollup-as-a-Service (RaaS) platform, currently powers over fifty EVM rollups (including Manta, Injective inEVM, Kinto, HYCHAIN, Treasure, and ApeChain) that secure ~$1B in total value locked. Along with their core rollup infrastructure, Caldera offers a catalog of over 50 integrations across the modular stack and an assortment of other custom user-facing tools for chains to leverage. Guardian Nodes are the latest addition to that list. The Background Today, the majority of L1 blockchains operate using a proof-of-stake model, where a chain’s participants are incentivized via network rewards to lend their stake and validate the chain. This incentive structure fosters decentralization by encouraging individual stakers to operate honest validators, while punishing those who don't, making for an extremely high cost to attack a network and compromise its security. Decentralizing Rollups EVM rollups have all but solved the scalability problem, with L2s and L3s on Ethereum enabling virtually infinite scale. But, there's still no incentive for honest network participants to monitor these rollups. Caldera’s Guardian Nodes allow teams to decentralize their rollups by enabling users to verify blocks and secure the network in exchange for rewards. Under the hood, this is accomplished by introducing a novel “light verifier'' to Arbitrum rollups that allows Guardian Node operators to verify Nitro batches on everyday hardware without needing to run a full node. Teams can launch Guardian Nodes to their users through a “Node Sale”, which distributes “keys” that authenticate a node's eligibility to submit claims and earn rewards, granting purchasers the ability to operate a Guardian Node on a given rollup. HYCHAIN, the first team to leverage this system, raised over $8m across 16,000+ node keys in just 2 weeks, completely supercharging their community while generating significant revenue for their project. By enabling more parties to watch over a rollup and identify malicious behavior, the network’s security grows more robust— a crucial step to establishing trust in the chain’s correctness. This in turn generates more demand for a rollup’s native token, which is required for users to participate in validation and helps provide practical cryptoeconomic security for the network. Guardians Nodes are another notch in the belt of Caldera’s impressive infrastructure solution. With the success of HYCHAIN's launch, we expect more teams building rollups to leverage this innovative product in the coming months. To get started with a high performance rollup, visit Caldera’s website here and/or book a call here. Contact Growth Alex Gu Caldera alex@caldera.xyz 4844329611 This story was distributed as a release by Chainwire under HackerNoon’s Business Blogging Program. Learn more about the program here. SAN FRANSCISCO, United States, May 24th, 2024/Chainwire/ SAN FRANSCISCO, United States, May 24th, 2024/Chainwire/ Rollup-as-a-Service platform Caldera is launching Guardian Nodes, a production-ready node system that enables network users to verify rollup blocks for rewards. HYCHAIN, the first team to leverage this system with Caldera, generated $8 million from their initial node sale. Rollup-as-a-Service platform Caldera is launching Guardian Nodes, a production-ready node system that enables network users to verify rollup blocks for rewards. HYCHAIN, the first team to leverage this system with Caldera, generated $8 million from their initial node sale. Caldera, the leading Rollup-as-a-Service (RaaS) platform, currently powers over fifty EVM rollups (including Manta, Injective inEVM, Kinto, HYCHAIN, Treasure, and ApeChain) that secure ~$1B in total value locked. Along with their core rollup infrastructure, Caldera offers a catalog of over 50 integrations across the modular stack and an assortment of other custom user-facing tools for chains to leverage. Guardian Nodes are the latest addition to that list. The Background Today, the majority of L1 blockchains operate using a proof-of-stake model, where a chain’s participants are incentivized via network rewards to lend their stake and validate the chain. This incentive structure fosters decentralization by encouraging individual stakers to operate honest validators, while punishing those who don't, making for an extremely high cost to attack a network and compromise its security. Decentralizing Rollups EVM rollups have all but solved the scalability problem, with L2s and L3s on Ethereum enabling virtually infinite scale. But, there's still no incentive for honest network participants to monitor these rollups. Caldera’s Guardian Nodes allow teams to decentralize their rollups by enabling users to verify blocks and secure the network in exchange for rewards. Under the hood, this is accomplished by introducing a novel “light verifier'' to Arbitrum rollups that allows Guardian Node operators to verify Nitro batches on everyday hardware without needing to run a full node. Teams can launch Guardian Nodes to their users through a “Node Sale”, which distributes “keys” that authenticate a node's eligibility to submit claims and earn rewards, granting purchasers the ability to operate a Guardian Node on a given rollup. HYCHAIN , the first team to leverage this system, raised over $8m across 16,000+ node keys in just 2 weeks, completely supercharging their community while generating significant revenue for their project. HYCHAIN HYCHAIN By enabling more parties to watch over a rollup and identify malicious behavior, the network’s security grows more robust— a crucial step to establishing trust in the chain’s correctness. This in turn generates more demand for a rollup’s native token, which is required for users to participate in validation and helps provide practical cryptoeconomic security for the network. Guardians Nodes are another notch in the belt of Caldera’s impressive infrastructure solution. With the success of HYCHAIN's launch, we expect more teams building rollups to leverage this innovative product in the coming months. To get started with a high performance rollup, visit Caldera’s website here and/or book a call here . here here here here Contact Growth Alex Gu Caldera alex@caldera.xyz 4844329611 This story was distributed as a release by Chainwire under HackerNoon’s Business Blogging Program. Learn more about the program here. This story was distributed as a release by Chainwire under HackerNoon’s Business Blogging Program. Learn more about the program here . here here

Caldera Launches Guardian Nodes, Enabling Teams To Raise Funds And Decentralize Their Network

As the founder of , an NFT project built on the Ethereum blockchain to empower women in the fight against cancer, I find myself mostly writing for our weekly newsletter, TFW Talks Cancer. Tutti Frutti Women Today, I want to extend an invitation to all cancer warriors, survivors, and those passionate about cancer awareness to stay updated and join our community by subscribing to our newsletter. If you're looking to stay informed, inspired, and connected in the world of cancer awareness, our weekly newsletter, TFW Talks Cancer, is the perfect platform for you. By subscribing to our newsletter, you'll receive valuable insights, stories, and resources related to cancer awareness directly to your inbox. Don't miss out on the opportunity to be part of this transformative movement. Subscribe to TFW Talks Cancer here: https://tuttifruttiwomen.substack.com/ In the month of May, we embarked on an incredible journey of cancer awareness, dedicating our newsletters to shedding light on different types of cancer. We explored topics such as brain cancer, bladder cancer, and melanoma and skin cancer, providing valuable information to our community. Our aim is to empower individuals with knowledge and create a supportive environment for those affected by cancer. As we venture into June, our commitment to cancer awareness continues to grow stronger. We have exciting plans and initiatives lined up to further educate and inspire our community. Here's a glimpse of what you can expect: National Cancer Survivors Day (June 4, 2023): Join us in celebrating the resilience and strength of cancer survivors worldwide. We'll be sharing inspiring stories and resources to uplift and support survivors on this special day. Men's Health Week (June 12-18, 2023): During this week, we'll be focusing on raising awareness about prostate cancer, testicular cancer, and penile cancer. Our goal is to promote open conversations and provide essential information to support men's health. World Kidney Cancer Day (June 15, 2023): Kidney cancer is a significant health concern, and we'll be dedicating this day to shed light on this topic. Stay tuned for educational content and resources to increase awareness and understanding of kidney cancer. National Black Family Cancer Awareness Week (June 15-21, 2023): We recognize the importance of addressing health disparities within different communities. During this week, we will emphasize the impact of cancer on the Black community and promote awareness and support for those affected. By subscribing to TFW Talks Cancer, you'll stay informed about these initiatives and more, ensuring you never miss an opportunity to contribute to the fight against cancer and support those who need it most. At Tutti Frutti Women, we believe in the power of community and knowledge-sharing when it comes to cancer awareness. Our weekly newsletter, TFW Talks Cancer, is a hub of valuable information, inspiring stories, and resources for individuals affected by cancer and those passionate about making a difference. Subscribe to our newsletter today and join our growing community of cancer warriors, survivors, and advocates. Together, we can create a future where cancer is understood, prevented, and conquered. Subscribe to TFW Talks Cancer here: https://tuttifruttiwomen.substack.com/ Stay informed, empowered, and connected in the world of cancer awareness.

Tutti Frutti Women: Empowering Cancer Awareness Through a Weekly Newsletter

Building Your AI Radiologist: A Fun Guide to Creating a Pneumonia Detector with VGG16

আপনার এআই রেডিওলজিস্ট তৈরি করা: VGG16 দিয়ে নিউমোনিয়া ডিটেক্টর তৈরির জন্য একটি মজার গাইড

Aufbau Ihres KI-Radiologen: Eine unterhaltsame Anleitung zum Erstellen eines Lungenentzündungsdetektors mit VGG16

Construyendo su radiólogo con IA: una guía divertida para crear un detector de neumonía con VGG16

Construire votre radiologue IA : un guide amusant pour créer un détecteur de pneumonie avec VGG16

अपना AI रेडियोलॉजिस्ट तैयार करें: VGG16 के साथ न्यूमोनिया डिटेक्टर बनाने के लिए एक मजेदार गाइड

AI 放射線科医の育成: VGG16 を使った肺炎検出器の作成ガイド

AI 방사선 전문의 구축: VGG16을 사용하여 폐렴 감지기 생성을 위한 재미있는 가이드

Construindo seu radiologista de IA: um guia divertido para criar um detector de pneumonia com VGG16

Создание вашего искусственного радиолога: забавное руководство по созданию детектора пневмонии с помощью VGG16

Yapay Zeka Radyoloğunuzu Oluşturma: VGG16 ile Pnömoni Dedektörü Oluşturmak İçin Eğlenceli Bir Kılavuz

Xây dựng bác sĩ X quang AI của bạn: Hướng dẫn thú vị để tạo máy phát hiện bệnh viêm phổi với VGG16

打造你的 AI 放射科医生：使用 VGG16 创建肺炎检测器的有趣指南

Does the U.S. President Have Full Immunity? Here's What the U.S. Supreme Court Had to Say

Survey and Analysis of Smart Contract Quality Assurance: Conclusions, Acknowledgement and References

"AI can enhance data collection, analysis, prediction, and evaluation processes"

"এআই ডেটা সংগ্রহ, বিশ্লেষণ, ভবিষ্যদ্বাণী এবং মূল্যায়ন প্রক্রিয়াগুলিকে উন্নত করতে পারে"

„KI kann die Prozesse der Datenerfassung, -analyse, -vorhersage und -auswertung verbessern“

"La IA puede mejorar los procesos de recopilación, análisis, predicción y evaluación de datos"

"L'IA peut améliorer les processus de collecte, d'analyse, de prédiction et d'évaluation de données"

"एआई डेटा संग्रह, विश्लेषण, भविष्यवाणी और मूल्यांकन प्रक्रियाओं को बढ़ा सकता है"

「AIはデータの収集、分析、予測、評価のプロセスを強化できます」

"AI는 데이터 수집, 분석, 예측, 평가 프로세스를 향상시킬 수 있습니다"

"A IA pode aprimorar os processos de coleta, análise, previsão e avaliação de dados"

«ИИ может улучшить процессы сбора, анализа, прогнозирования и оценки данных»

"Yapay zeka veri toplama, analiz, tahmin ve değerlendirme süreçlerini geliştirebilir"

"AI có thể tăng cường quá trình thu thập, phân tích, dự đoán và đánh giá dữ liệu"

“人工智能可以增强数据收集、分析、预测和评估过程”

Router Protocol is pleased to announce that it has completed an oversubscribed strategic funding round with participation from prominent crypto angels and funds. The primary goal of the strategic round was to thoughtfully expand Router's cap table with crypto operators who deeply understand the importance of cross-chain infrastructure and can accelerate the ecosystem growth as Router evolves into a Layer 1 blockchain. Router Protocol is honored to welcome an exceptional group of accomplished founders and web3 executives as angels and advisors, including Keone Hon (Monad), Parker Jou (Caldera), Amrit Kumar (Altlayer), Aniket (Biconomy), Anurag Arjun (Avail), Amitej (Stader), Ravindra (Frontier), and several others. In a testament to the unwavering belief in its vision, Router Protocol also received overwhelming support from existing investors, including heavyweight players such as Wintermute Ventures, Woodstock, DeFi Capital, CryptoBanter, RavenDAO, and Luganodes. Other renowned advisors and angels include Zaki Manian (Sommelier/Cosmos), Ajit Tripathi (@chainyoda), Gokul Rajaram (Google/Doordash), and Surojit Chatterjee (Coinbase, Google, Flipkart, EMA). The proceeds from this funding round will be deployed strategically to fuel the development and highly anticipated launch of the Router Chain mainnet. Router Chain is set to revolutionize value transfer by abstracting away blockchain complexity and unlocking a new era of frictionless cross-chain applications and user experiences. "Router Chain embodies the evolution of our unwavering vision to build the universal interconnectivity layer between fragmented L1/L2 networks," said Ramani Ramachandran, CEO of Router Protocol. "By abstracting away blockchain complexity, we are laser-focused on delivering the most seamless and intuitive user and developer experience for multi-chain applications. This funding round propels us closer to realising this transformative vision." Router Protocol’s mantra has always been "developers first." Its mission is to empower builders to focus on bringing their groundbreaking ideas to life, rather than grappling with infrastructure complexities. Router Chain will provide a robust suite of flexible interoperability primitives, including its innovative cross-chain intent framework and developer-friendly toolkits, to make this vision a reality. "We are thrilled to witness the rapid expansion of our ecosystem, with trailblazing cross-chain projects like StakeEase, FolioX, and many more leveraging our interoperability primitives to push the boundaries of what's possible," said Shubham Singh, CTO of Router Protocol. In a major milestone, Router Protocol launched Router Nitro, a super-fast, gas-efficient, intent-based cross-chain bridge powered by an innovative reverse-verification mechanism. Nitro Bridge crossed $350M volume in the past 4 months, over 650K+ transactions, and almost 300K unique users. To further accelerate ecosystem growth, Router Protocol is thrilled to announce the upcoming launch of the Router Ecosystem Grants Program, which will provide substantial financial support and resources to promising projects building on Router Chain. Moreover, the project’s recently revamped tokenomics empowers it to aggressively focus on community engagement, ecosystem expansion, and the promotion of exciting new cross-chain primitives that will redefine the landscape of interoperability. The oversubscribed round demonstrates the demand for high-value web3 infrastructure and marks a crucial step in Router's maturation from a bridge solution to a core Web3 connecting layer. Backing from this expanded group of partners will aid in evangelizing Router's vision and unique value proposition to a global audience. Router Protocol is pleased to announce that it has completed an oversubscribed strategic funding round with participation from prominent crypto angels and funds. Router Protocol is pleased to announce that it has completed an oversubscribed strategic funding round with participation from prominent crypto angels and funds. The primary goal of the strategic round was to thoughtfully expand Router's cap table with crypto operators who deeply understand the importance of cross-chain infrastructure and can accelerate the ecosystem growth as Router evolves into a Layer 1 blockchain. Router Protocol is honored to welcome an exceptional group of accomplished founders and web3 executives as angels and advisors, including Keone Hon (Monad), Parker Jou (Caldera), Amrit Kumar (Altlayer), Aniket (Biconomy), Anurag Arjun (Avail), Amitej (Stader), Ravindra (Frontier), and several others. In a testament to the unwavering belief in its vision, Router Protocol also received overwhelming support from existing investors, including heavyweight players such as Wintermute Ventures, Woodstock, DeFi Capital, CryptoBanter, RavenDAO, and Luganodes. Other renowned advisors and angels include Zaki Manian (Sommelier/Cosmos), Ajit Tripathi (@chainyoda), Gokul Rajaram (Google/Doordash), and Surojit Chatterjee (Coinbase, Google, Flipkart, EMA). The proceeds from this funding round will be deployed strategically to fuel the development and highly anticipated launch of the Router Chain mainnet. Router Chain is set to revolutionize value transfer by abstracting away blockchain complexity and unlocking a new era of frictionless cross-chain applications and user experiences. "Router Chain embodies the evolution of our unwavering vision to build the universal interconnectivity layer between fragmented L1/L2 networks," said Ramani Ramachandran, CEO of Router Protocol. "By abstracting away blockchain complexity, we are laser-focused on delivering the most seamless and intuitive user and developer experience for multi-chain applications. This funding round propels us closer to realising this transformative vision." Router Protocol’s mantra has always been "developers first." Its mission is to empower builders to focus on bringing their groundbreaking ideas to life, rather than grappling with infrastructure complexities. Router Chain will provide a robust suite of flexible interoperability primitives, including its innovative cross-chain intent framework and developer-friendly toolkits, to make this vision a reality. "We are thrilled to witness the rapid expansion of our ecosystem, with trailblazing cross-chain projects like StakeEase, FolioX, and many more leveraging our interoperability primitives to push the boundaries of what's possible," said Shubham Singh, CTO of Router Protocol. In a major milestone, Router Protocol launched Router Nitro, a super-fast, gas-efficient, intent-based cross-chain bridge powered by an innovative reverse-verification mechanism. Nitro Bridge crossed $350M volume in the past 4 months, over 650K+ transactions, and almost 300K unique users. To further accelerate ecosystem growth, Router Protocol is thrilled to announce the upcoming launch of the Router Ecosystem Grants Program, which will provide substantial financial support and resources to promising projects building on Router Chain. Moreover, the project’s recently revamped tokenomics empowers it to aggressively focus on community engagement, ecosystem expansion, and the promotion of exciting new cross-chain primitives that will redefine the landscape of interoperability. The oversubscribed round demonstrates the demand for high-value web3 infrastructure and marks a crucial step in Router's maturation from a bridge solution to a core Web3 connecting layer. Backing from this expanded group of partners will aid in evangelizing Router's vision and unique value proposition to a global audience.

Router Protocol Completes Oversubscribed Strategic Funding Round

রাউটার প্রোটোকল ওভারসাবস্ক্রাইবড স্ট্র্যাটেজিক ফান্ডিং রাউন্ড সম্পূর্ণ করে

Router Protocol schließt überzeichnete strategische Finanzierungsrunde ab

Router Protocol completa una ronda de financiación estratégica con exceso de suscripción

Le protocole de routeur finalise un cycle de financement stratégique sursouscrit

राउटर प्रोटोकॉल ने ओवरसब्सक्राइब्ड रणनीतिक फंडिंग राउंड पूरा किया

ルータープロトコル、募集額超過の戦略的資金調達ラウンドを完了

라우터 프로토콜, 초과 구독된 전략적 자금 조달 라운드 완료

Router Protocol conclui rodada de financiamento estratégico com excesso de inscrições

Router Protocol завершает раунд стратегического финансирования переподписки

Yönlendirici Protokolü Aşırı Abone Olunan Stratejik Finansman Turunu Tamamladı

Giao thức bộ định tuyến hoàn thành vòng tài trợ chiến lược được đăng ký vượt mức

路由器协议完成超额认购的战略融资轮

Authors:
(1) Michael Moor, Department of Computer Science, Stanford University, Stanford, USA and these authors contributed equally to this work;
(2) Qian Huang, Department of Computer Science, Stanford University, Stanford, USA and these authors contributed equally to this work;
(3) Shirley Wu, Department of Computer Science, Stanford University, Stanford, USA;
(4) Michihiro Yasunaga, Department of Computer Science, Stanford University, Stanford, USA;
(5) Cyril Zakka, Department of Cardiothoracic Surgery, Stanford Medicine, Stanford, USA;
(6) Yash Dalmia, Department of Computer Science, Stanford University, Stanford, USA;
(7) Eduardo Pontes Reis, Hospital Israelita Albert Einstein, Sao Paulo, Brazil;
(8) Pranav Rajpurkar, Department of Biomedical Informatics, Harvard Medical School, Boston, USA;
(9) Jure Leskovec, Department of Computer Science, Stanford University, Stanford, USA. Table of Links Abstract and 1 Introduction 2 Related Works 3 Med-Flamingo 4 Evaluation 5 Results 6 Discussion, Acknowledgments, and References A Appendix 5 RESULTS In our experiments, we focus on generative medical visual question answering (VQA). While recent medical VLMs predominantly performed VQA in a non-generative but rather discriminative manner (i.e., by scoring different answer choices), we believe that this ex-post classification to carry less clinical usefulness, than directly generating responses. On the other hand, generative VQA is more challenging to evaluate, as automated metrics suffer from significant limitations as they do not fully capture the domain-specific context. Thus, we perform a human evaluation study where clinical experts review model generations (blinded) and score them (between 0 and 10) in terms of clinical usefulness. Conventional VQA datasets Table 1 shows the results for VQA-RAD, the radiological VQA dataset for which we created custom splits to address leakage (see Section4). Med-Flamingo few-shot shows strong results, improving the clinical eval score by ∼ 20% over the best baseline. In this dataset, the auxiliary metrics are rather aligned with clinical preference. Finetuning the MedVINT baseline did not lead to improved performance on this dataset which may be due to its small size. MedVINT zero-shot outperforms the other zero-shot ablations which may be partially attributed to its instruction tuning step on PMC-VQA. Table 2 shows for the results for Path-VQA, the pathology VQA dataset. Compared to the other datasets, all models overall perform poorer on the Path-VQA dataset in terms of clinical evaluation score. We hypothesize that this has to do with the fact the models are not pre-trained on actual large-scale and fine-grained pathology image datasets, but only on a rather small amount of pathology literature (which may not be enough to achieve strong performance). For instance, Figure 3 shows that only a small fraction of our training data covers pathology. In the automated metrics (BERT-sim and exact-match), Med-Flamingo improves upon the OpenFlamingo baseline, however the overall quality does not improve (as seen in the clinical evaluation score). MedVINT was fine-tuned on a sizeable training split which results in strong automated metrics, but did not result in a clinical evaluation score that matches any Flamingo variant. Visual USMLE Table 3 shows the results for the Visual USMLE dataset. Med-Flamingo (fewshot) results in the clinically most preferrable generations, whereas OpenFlamingo (zero-shot) is a close runner-up. As the ground truth answers were rather lengthy paragraphs, exact match was not an informative metric (constant 0 for all methods). The few-shot prompted models lead to lower automated scores than their zero-shot counterparts, which we hypothesize has to do with the fact that the USMLE problems are long (long vignettes as well as long answers) which forced us to summarize the questions and answers when designing few-shot prompts (for which we used GPT-4). Hence, it’s possible that those prompts lead to short answers that in terms of BERT-sim score may differ more from the correct answer than a more wordy zero-shot generation. Across datasets Overall, we find that Med-Flamingo’s multimodal in-domain few-shot learning abilities lead to favorable generative VQA performance, leading to the lowest average rank of 1.67 in terms of clinical evaluation score as averaged across all evaluation datasets. As runner-up, OpenFlamingo zero-shot achieves a rank of 2.33. Qualitative analysis Finally, we showcase few examples of Med-Flamingo generations in more detail in Figures 1,5, and 6. Figure 5 exemplifies that a medical few-shot learner like Med-Flamingo can be prompted to generate rationale for its VQA answer. The shown example is impressive in that the rationale is visually guiding the reader towards the object of interest (calcification of the aortic wall). We note, however, that at this stage, few-shot multimodal prompted rationales may not be robust, especially when a model arrives at a wrong answer. Figures 1 and 6 showcase two example problems from the Visual USMLE dataset. The problem descriptions were slightly rephrased and summarized using GPT-4 for display. In Figure 6, MedFlamingo generates the correct answer while not mentioning the underlying diagnosis (urothelial cancer) as it was not asked for. By contrast, we observed baselines to directly diagnose the patient (instead of answering the actual question in a targeted way). The problem in Figure 1 illustrates that Med-Flamingo has the ability to integrate complex medical history information together with visual information to synthesize a comprehensive diagnosis that draws from the information of both modalities. This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. Authors: (1) Michael Moor, Department of Computer Science, Stanford University, Stanford, USA and these authors contributed equally to this work; (2) Qian Huang, Department of Computer Science, Stanford University, Stanford, USA and these authors contributed equally to this work; (3) Shirley Wu, Department of Computer Science, Stanford University, Stanford, USA; (4) Michihiro Yasunaga, Department of Computer Science, Stanford University, Stanford, USA; (5) Cyril Zakka, Department of Cardiothoracic Surgery, Stanford Medicine, Stanford, USA; (6) Yash Dalmia, Department of Computer Science, Stanford University, Stanford, USA; (7) Eduardo Pontes Reis, Hospital Israelita Albert Einstein, Sao Paulo, Brazil; (8) Pranav Rajpurkar, Department of Biomedical Informatics, Harvard Medical School, Boston, USA; (9) Jure Leskovec, Department of Computer Science, Stanford University, Stanford, USA. Authors: Authors: (1) Michael Moor, Department of Computer Science, Stanford University, Stanford, USA and these authors contributed equally to this work; (2) Qian Huang, Department of Computer Science, Stanford University, Stanford, USA and these authors contributed equally to this work; (3) Shirley Wu, Department of Computer Science, Stanford University, Stanford, USA; (4) Michihiro Yasunaga, Department of Computer Science, Stanford University, Stanford, USA; (5) Cyril Zakka, Department of Cardiothoracic Surgery, Stanford Medicine, Stanford, USA; (6) Yash Dalmia, Department of Computer Science, Stanford University, Stanford, USA; (7) Eduardo Pontes Reis, Hospital Israelita Albert Einstein, Sao Paulo, Brazil; (8) Pranav Rajpurkar, Department of Biomedical Informatics, Harvard Medical School, Boston, USA; (9) Jure Leskovec, Department of Computer Science, Stanford University, Stanford, USA. Table of Links Abstract and 1 Introduction Abstract and 1 Introduction 2 Related Works 2 Related Works 3 Med-Flamingo 3 Med-Flamingo 4 Evaluation 4 Evaluation 5 Results 5 Results 6 Discussion, Acknowledgments, and References 6 Discussion, Acknowledgments, and References A Appendix A Appendix 5 RESULTS In our experiments, we focus on generative medical visual question answering (VQA). While recent medical VLMs predominantly performed VQA in a non-generative but rather discriminative manner (i.e., by scoring different answer choices), we believe that this ex-post classification to carry less clinical usefulness, than directly generating responses. On the other hand, generative VQA is more challenging to evaluate, as automated metrics suffer from significant limitations as they do not fully capture the domain-specific context. Thus, we perform a human evaluation study where clinical experts review model generations (blinded) and score them (between 0 and 10) in terms of clinical usefulness. Conventional VQA datasets Table 1 shows the results for VQA-RAD, the radiological VQA dataset for which we created custom splits to address leakage (see Section4). Med-Flamingo few-shot shows strong results, improving the clinical eval score by ∼ 20% over the best baseline. In this dataset, the auxiliary metrics are rather aligned with clinical preference. Finetuning the MedVINT baseline did not lead to improved performance on this dataset which may be due to its small size. MedVINT zero-shot outperforms the other zero-shot ablations which may be partially attributed to its instruction tuning step on PMC-VQA. Conventional VQA datasets Table 2 shows for the results for Path-VQA, the pathology VQA dataset. Compared to the other datasets, all models overall perform poorer on the Path-VQA dataset in terms of clinical evaluation score. We hypothesize that this has to do with the fact the models are not pre-trained on actual large-scale and fine-grained pathology image datasets, but only on a rather small amount of pathology literature (which may not be enough to achieve strong performance). For instance, Figure 3 shows that only a small fraction of our training data covers pathology. In the automated metrics (BERT-sim and exact-match), Med-Flamingo improves upon the OpenFlamingo baseline, however the overall quality does not improve (as seen in the clinical evaluation score). MedVINT was fine-tuned on a sizeable training split which results in strong automated metrics, but did not result in a clinical evaluation score that matches any Flamingo variant. Visual USMLE Table 3 shows the results for the Visual USMLE dataset. Med-Flamingo (fewshot) results in the clinically most preferrable generations, whereas OpenFlamingo (zero-shot) is a close runner-up. As the ground truth answers were rather lengthy paragraphs, exact match was not an informative metric (constant 0 for all methods). The few-shot prompted models lead to lower automated scores than their zero-shot counterparts, which we hypothesize has to do with the fact that the USMLE problems are long (long vignettes as well as long answers) which forced us to summarize the questions and answers when designing few-shot prompts (for which we used GPT-4). Hence, it’s possible that those prompts lead to short answers that in terms of BERT-sim score may differ more from the correct answer than a more wordy zero-shot generation. Visual USMLE Across datasets Overall, we find that Med-Flamingo’s multimodal in-domain few-shot learning abilities lead to favorable generative VQA performance, leading to the lowest average rank of 1.67 in terms of clinical evaluation score as averaged across all evaluation datasets. As runner-up, OpenFlamingo zero-shot achieves a rank of 2.33. Across datasets Qualitative analysis Finally, we showcase few examples of Med-Flamingo generations in more detail in Figures 1,5, and 6. Figure 5 exemplifies that a medical few-shot learner like Med-Flamingo can be prompted to generate rationale for its VQA answer. The shown example is impressive in that the rationale is visually guiding the reader towards the object of interest (calcification of the aortic wall). We note, however, that at this stage, few-shot multimodal prompted rationales may not be robust, especially when a model arrives at a wrong answer. Qualitative analysis Figures 1 and 6 showcase two example problems from the Visual USMLE dataset. The problem descriptions were slightly rephrased and summarized using GPT-4 for display. In Figure 6, MedFlamingo generates the correct answer while not mentioning the underlying diagnosis (urothelial cancer) as it was not asked for. By contrast, we observed baselines to directly diagnose the patient (instead of answering the actual question in a targeted way). The problem in Figure 1 illustrates that Med-Flamingo has the ability to integrate complex medical history information together with visual information to synthesize a comprehensive diagnosis that draws from the information of both modalities. This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. available on arxiv

Med-Flamingo: a Multimodal Medical Few-shot Learner - Results

How Do You Go About Creating Biological Weapons That Cannot Be Countered?

Deep Neural Networks to Detect and Quantify Lymphoma Lesions: Results

ডিপ নিউরাল নেটওয়ার্ক লিম্ফোমা ক্ষত সনাক্ত এবং পরিমাপ করতে: ফলাফল

Tiefe neuronale Netzwerke zur Erkennung und Quantifizierung von Lymphomläsionen: Ergebnisse

Redes neuronales profundas para detectar y cuantificar lesiones de linfoma: resultados

Réseaux de neurones profonds pour détecter et quantifier les lésions de lymphome : résultats

लिम्फोमा घावों का पता लगाने और मात्रा निर्धारित करने के लिए डीप न्यूरल नेटवर्क: परिणाम

リンパ腫病変を検出し定量化するディープニューラルネットワーク：結果

림프종 병변을 탐지하고 정량화하기 위한 심층 신경망: 결과

Redes Neurais Profundas para Detectar e Quantificar Lesões de Linfoma: Resultados

Глубокие нейронные сети для обнаружения и количественной оценки поражений лимфомой: результаты

Lenfoma Lezyonlarını Tespit Etmek ve Ölçmek İçin Derin Sinir Ağları: Sonuçlar

Mạng lưới thần kinh sâu để phát hiện và định lượng tổn thương ung thư hạch: Kết quả

深度神经网络检测和量化淋巴瘤病变：结果

Deep Neural Networks to Detect and Quantify Lymphoma Lesions: Materials and Methods

ডিপ নিউরাল নেটওয়ার্ক লিম্ফোমা ক্ষত সনাক্ত এবং পরিমাপ করার জন্য: উপাদান এবং পদ্ধতি

Tiefe neuronale Netzwerke zur Erkennung und Quantifizierung von Lymphomläsionen: Materialien und Methoden

Redes neuronales profundas para detectar y cuantificar lesiones de linfoma: materiales y métodos

Réseaux de neurones profonds pour détecter et quantifier les lésions de lymphome : matériels et méthodes

लिम्फोमा घावों का पता लगाने और मात्रा निर्धारित करने के लिए डीप न्यूरल नेटवर्क: सामग्री और विधियाँ

リンパ腫病変を検出し定量化するディープニューラルネットワーク：材料と方法

림프종 병변을 검출하고 정량화하기 위한 심층 신경망: 재료 및 방법

Redes Neurais Profundas para Detectar e Quantificar Lesões de Linfoma: Materiais e Métodos

Глубокие нейронные сети для обнаружения и количественной оценки поражений лимфомой: материалы и методы

Lenfoma Lezyonlarını Tespit Etmek ve Ölçmek İçin Derin Sinir Ağları: Malzemeler ve Yöntemler

Mạng lưới thần kinh sâu để phát hiện và định lượng tổn thương ung thư hạch: Vật liệu và phương pháp

深度神经网络检测和量化淋巴瘤病变：材料和方法

Authors:
(1) Shadab Ahamed, University of British Columbia, Vancouver, BC, Canada, BC Cancer Research Institute, Vancouver, BC, Canada. He was also a Mitacs Accelerate Fellow (May 2022 - April 2023) with Microsoft AI for Good Lab, Redmond, WA, USA (e-mail: shadabahamed1996@gmail.com);
(2) Yixi Xu, Microsoft AI for Good Lab, Redmond, WA, USA;
(3) Claire Gowdy, BC Children’s Hospital, Vancouver, BC, Canada;
(4) Joo H. O, St. Mary’s Hospital, Seoul, Republic of Korea;
(5) Ingrid Bloise, BC Cancer, Vancouver, BC, Canada;
(6) Don Wilson, BC Cancer, Vancouver, BC, Canada;
(7) Patrick Martineau, BC Cancer, Vancouver, BC, Canada;
(8) Franc¸ois Benard, BC Cancer, Vancouver, BC, Canada;
(9) Fereshteh Yousefirizi, BC Cancer Research Institute, Vancouver, BC, Canada;
(10) Rahul Dodhia, Microsoft AI for Good Lab, Redmond, WA, USA;
(11) Juan M. Lavista, Microsoft AI for Good Lab, Redmond, WA, USA;
(12) William B. Weeks, Microsoft AI for Good Lab, Redmond, WA, USA;
(13) Carlos F. Uribe, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada;
(14) Arman Rahmim, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada. Table of Links Abstract and Intro
Related Work
Materials and Methods
Results
Discussion
Conclusion and References VI. CONCLUSION In this study, we assessed various neural network architectures for automating lymphoma lesion segmentation in PET/CT images across multiple datasets. We examined the reproducibility of lesion measures, revealing differences among networks, highlighting their suitability for specific clinical uses. Additionally, we introduced three lesion detection criteria to assess network performance at a per-lesion level, emphasizing their clinical relevance. Lastly, we discussed challenges related to ground truth consistency and stressed the importance of having well-defined protocol for segmentation. This work provides valuable insights into deep learning’s potentials and limitations in lymphoma lesion segmentation and emphasizes the need for standardized annotation practices to enhance research validity and clinical applications. REFERENCES [1] S. F. Barrington et al. “FDG PET for therapy monitoring in Hodgkin and non-Hodgkin lymphomas”. In: European Journal of Nuclear Medicine and Molecular Imaging 44.1 (Aug. 2017), pp. 97–110. ISSN: 1619- 7089. [2] K. Okuyucu et al. “Prognosis estimation under the light of metabolic tumor parameters on initial FDG-PET/CT in patients with primary extranodal lymphoma”. en. In: Radiol. Oncol. 50.4 (Dec. 2016), pp. 360–369. [3] N. Wu et al. “Deep Neural Networks Improve Radiologists’ Performance in Breast Cancer Screening”. In: IEEE Transactions on Medical Imaging 39.4 (2020), pp. 1184–1194. [4] C. Yuan et al. “Diffuse large B-cell lymphoma segmentation in PET-CT images via hybrid learning for feature fusion”. In: Medical Physics 48.7 (2021), pp. 3665– 3678. [5] H. Hu et al. “Lymphoma Segmentation in PET Images Based on Multi-view and Conv3D Fusion Strategy”. In: 2020 IEEE 17th International Symposium on Biomedical Imaging (ISBI). 2020, pp. 1197–1200. [6] H. Li et al. “DenseX-Net: An End-to-End Model for Lymphoma Segmentation in Whole-Body PET/CT Images”. In: IEEE Access 8 (2020), pp. 8004–8018. [7] L. Liu et al. “Improved Multi-modal Patch Based Lymphoma Segmentation with Negative Sample Augmentation and Label Guidance on PET/CT Scans”. In: Multiscale Multimodal Medical Imaging. Ed. by X. Li et al. Cham: Springer Nature Switzerland, 2022, pp. 121–129. ISBN: 978-3-031-18814-5. [8] C. S. Constantino et al. “Evaluation of Semiautomatic and Deep Learning–Based Fully Automatic Segmentation Methods on [18F]FDG PET/CT Images from Patients with Lymphoma: Influence on Tumor Characterization”. In: Journal of Digital Imaging 36.4 (Aug. 2023), pp. 1864–1876. ISSN: 1618-727X. [9] A. J. Weisman et al. “Comparison of 11 automated PET segmentation methods in lymphoma”. English. In: Physics in medicine & biology 65.23 (2020), pp. 235019–235019. [10] A. J. Weisman et al. “Convolutional Neural Networks for Automated PET/CT Detection of Diseased Lymph Node Burden in Patients with Lymphoma”. In: Radiology: Artificial Intelligence 2.5 (2020), e200016. [11] C. Jiang et al. “Deep learning–based tumour segmentation and total metabolic tumour volume prediction in the prognosis of diffuse large B-cell lymphoma patients in 3D FDG-PET images”. In: European Radiology 32.7 (July 2022), pp. 4801–4812. ISSN: 1432-1084. [12] P. Blanc-Durand et al. “Fully automatic segmentation of diffuse large B cell lymphoma lesions on 3D FDGPET/CT for total metabolic tumour volume prediction using a convolutional neural network.” In: European Journal of Nuclear Medicine and Molecular Imaging 48.5 (May 2021), pp. 1362–1370. ISSN: 1619-7089. [13] S. Ahamed et al. “A U-Net Convolutional Neural Network with Multiclass Dice Loss for Automated Segmentation of Tumors and Lymph Nodes from Head and Neck Cancer PET/CT Images”. In: Head and Neck Tumor Segmentation and Outcome Prediction. Ed. by V. Andrearczyk et al. Cham: Springer Nature Switzerland, 2023, pp. 94–106. ISBN: 978-3-031-27420-6. [14] S. Gatidis et al. “A whole-body FDG-PET/CT Dataset with manually annotated Tumor Lesions”. In: Scientific Data 9.1 (Oct. 2022), p. 601. ISSN: 2052-4463. [15] M. Pop et al. “Left-Ventricle Quantification Using Residual U-Net”. English. In: vol. 11395. Statistical Atlases and Computational Models of the Heart. Atrial Segmentation and LV Quantification Challenges. Switzerland: Springer International Publishing AG, 2019, pp. 371–380. ISBN: 0302-9743. [16] A. Myronenko. “3D MRI Brain Tumor Segmentation Using Autoencoder Regularization”. English. In: Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries. Cham: Springer International Publishing, 2019, pp. 311–320. ISBN: 0302-9743. [17] F. Isensee et al. “nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation”. In: Nature Methods 18.2 (Dec. 2020), pp. 203–211. [18] A. Hatamizadeh et al. “Swin UNETR: Swin Transformers for Semantic Segmentation of Brain Tumors in MRI Images”. English. In: (2022). [19] M. J. Cardoso et al. MONAI: An open-source framework for deep learning in healthcare. 2022. arXiv: 2211. 02701 [cs.LG]. [20] S. Ahamed et al. “Towards enhanced lesion segmentation using a 3D neural network trained on multiresolution cropped patches of lymphoma PET images”. In: Journal of Nuclear Medicine 64.supplement 1 (2023), P1360–P1360. ISSN: 0161-5505. [21] J. L. Fleiss. “Measuring nominal scale agreement among many raters”. English. In: Psychological bulletin 76.5 (1971), pp. 378–382. [22] A. K. Jha et al. “Nuclear Medicine and artificial intelligence: Best practices for evaluation (the RELAINCE guidelines)”. en. In: J. Nucl. Med. 63.9 (Sept. 2022), pp. 1288–1299. [23] N. Hasani et al. “Artificial Intelligence in Lymphoma PET Imaging: A Scoping Review (Current Trends and Future Directions)”. In: PET clinics 17.1 (Jan. 1), pp. 145–174. ISSN: 1556-8598. [24] S. K. Warfield et al. “Simultaneous truth and performance level estimation (STAPLE): an algorithm for the validation of image segmentation”. en. In: IEEE Trans. Med. Imaging 23.7 (July 2004), pp. 903–921. This paper is available on arxiv under CC 4.0 license. Authors: (1) Shadab Ahamed, University of British Columbia, Vancouver, BC, Canada, BC Cancer Research Institute, Vancouver, BC, Canada. He was also a Mitacs Accelerate Fellow (May 2022 - April 2023) with Microsoft AI for Good Lab, Redmond, WA, USA (e-mail: shadabahamed1996@gmail.com); (2) Yixi Xu, Microsoft AI for Good Lab, Redmond, WA, USA; (3) Claire Gowdy, BC Children’s Hospital, Vancouver, BC, Canada; (4) Joo H. O, St. Mary’s Hospital, Seoul, Republic of Korea; (5) Ingrid Bloise, BC Cancer, Vancouver, BC, Canada; (6) Don Wilson, BC Cancer, Vancouver, BC, Canada; (7) Patrick Martineau, BC Cancer, Vancouver, BC, Canada; (8) Franc¸ois Benard, BC Cancer, Vancouver, BC, Canada; (9) Fereshteh Yousefirizi, BC Cancer Research Institute, Vancouver, BC, Canada; (10) Rahul Dodhia, Microsoft AI for Good Lab, Redmond, WA, USA; (11) Juan M. Lavista, Microsoft AI for Good Lab, Redmond, WA, USA; (12) William B. Weeks, Microsoft AI for Good Lab, Redmond, WA, USA; (13) Carlos F. Uribe, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada; (14) Arman Rahmim, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada. Authors: Authors: (1) Shadab Ahamed, University of British Columbia, Vancouver, BC, Canada, BC Cancer Research Institute, Vancouver, BC, Canada. He was also a Mitacs Accelerate Fellow (May 2022 - April 2023) with Microsoft AI for Good Lab, Redmond, WA, USA (e-mail: shadabahamed1996@gmail.com); (2) Yixi Xu, Microsoft AI for Good Lab, Redmond, WA, USA; (3) Claire Gowdy, BC Children’s Hospital, Vancouver, BC, Canada; (4) Joo H. O, St. Mary’s Hospital, Seoul, Republic of Korea; (5) Ingrid Bloise, BC Cancer, Vancouver, BC, Canada; (6) Don Wilson, BC Cancer, Vancouver, BC, Canada; (7) Patrick Martineau, BC Cancer, Vancouver, BC, Canada; (8) Franc¸ois Benard, BC Cancer, Vancouver, BC, Canada; (9) Fereshteh Yousefirizi, BC Cancer Research Institute, Vancouver, BC, Canada; (10) Rahul Dodhia, Microsoft AI for Good Lab, Redmond, WA, USA; (11) Juan M. Lavista, Microsoft AI for Good Lab, Redmond, WA, USA; (12) William B. Weeks, Microsoft AI for Good Lab, Redmond, WA, USA; (13) Carlos F. Uribe, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada; (14) Arman Rahmim, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada. Table of Links Abstract and Intro Related Work Materials and Methods Results Discussion Conclusion and References Abstract and Intro Abstract and Intro Related Work Related Work Materials and Methods Materials and Methods Results Results Discussion Discussion Conclusion and References Conclusion and References VI. CONCLUSION In this study, we assessed various neural network architectures for automating lymphoma lesion segmentation in PET/CT images across multiple datasets. We examined the reproducibility of lesion measures, revealing differences among networks, highlighting their suitability for specific clinical uses. Additionally, we introduced three lesion detection criteria to assess network performance at a per-lesion level, emphasizing their clinical relevance. Lastly, we discussed challenges related to ground truth consistency and stressed the importance of having well-defined protocol for segmentation. This work provides valuable insights into deep learning’s potentials and limitations in lymphoma lesion segmentation and emphasizes the need for standardized annotation practices to enhance research validity and clinical applications. REFERENCES [1] S. F. Barrington et al. “FDG PET for therapy monitoring in Hodgkin and non-Hodgkin lymphomas”. In: European Journal of Nuclear Medicine and Molecular Imaging 44.1 (Aug. 2017), pp. 97–110. ISSN: 1619- 7089. [2] K. Okuyucu et al. “Prognosis estimation under the light of metabolic tumor parameters on initial FDG-PET/CT in patients with primary extranodal lymphoma”. en. In: Radiol. Oncol. 50.4 (Dec. 2016), pp. 360–369. [3] N. Wu et al. “Deep Neural Networks Improve Radiologists’ Performance in Breast Cancer Screening”. In: IEEE Transactions on Medical Imaging 39.4 (2020), pp. 1184–1194. [4] C. Yuan et al. “Diffuse large B-cell lymphoma segmentation in PET-CT images via hybrid learning for feature fusion”. In: Medical Physics 48.7 (2021), pp. 3665– 3678. [5] H. Hu et al. “Lymphoma Segmentation in PET Images Based on Multi-view and Conv3D Fusion Strategy”. In: 2020 IEEE 17th International Symposium on Biomedical Imaging (ISBI). 2020, pp. 1197–1200. [6] H. Li et al. “DenseX-Net: An End-to-End Model for Lymphoma Segmentation in Whole-Body PET/CT Images”. In: IEEE Access 8 (2020), pp. 8004–8018. [7] L. Liu et al. “Improved Multi-modal Patch Based Lymphoma Segmentation with Negative Sample Augmentation and Label Guidance on PET/CT Scans”. In: Multiscale Multimodal Medical Imaging. Ed. by X. Li et al. Cham: Springer Nature Switzerland, 2022, pp. 121–129. ISBN: 978-3-031-18814-5. [8] C. S. Constantino et al. “Evaluation of Semiautomatic and Deep Learning–Based Fully Automatic Segmentation Methods on [18F]FDG PET/CT Images from Patients with Lymphoma: Influence on Tumor Characterization”. In: Journal of Digital Imaging 36.4 (Aug. 2023), pp. 1864–1876. ISSN: 1618-727X. [9] A. J. Weisman et al. “Comparison of 11 automated PET segmentation methods in lymphoma”. English. In: Physics in medicine & biology 65.23 (2020), pp. 235019–235019. [10] A. J. Weisman et al. “Convolutional Neural Networks for Automated PET/CT Detection of Diseased Lymph Node Burden in Patients with Lymphoma”. In: Radiology: Artificial Intelligence 2.5 (2020), e200016. [11] C. Jiang et al. “Deep learning–based tumour segmentation and total metabolic tumour volume prediction in the prognosis of diffuse large B-cell lymphoma patients in 3D FDG-PET images”. In: European Radiology 32.7 (July 2022), pp. 4801–4812. ISSN: 1432-1084. [12] P. Blanc-Durand et al. “Fully automatic segmentation of diffuse large B cell lymphoma lesions on 3D FDGPET/CT for total metabolic tumour volume prediction using a convolutional neural network.” In: European Journal of Nuclear Medicine and Molecular Imaging 48.5 (May 2021), pp. 1362–1370. ISSN: 1619-7089. [13] S. Ahamed et al. “A U-Net Convolutional Neural Network with Multiclass Dice Loss for Automated Segmentation of Tumors and Lymph Nodes from Head and Neck Cancer PET/CT Images”. In: Head and Neck Tumor Segmentation and Outcome Prediction. Ed. by V. Andrearczyk et al. Cham: Springer Nature Switzerland, 2023, pp. 94–106. ISBN: 978-3-031-27420-6. [14] S. Gatidis et al. “A whole-body FDG-PET/CT Dataset with manually annotated Tumor Lesions”. In: Scientific Data 9.1 (Oct. 2022), p. 601. ISSN: 2052-4463. [15] M. Pop et al. “Left-Ventricle Quantification Using Residual U-Net”. English. In: vol. 11395. Statistical Atlases and Computational Models of the Heart. Atrial Segmentation and LV Quantification Challenges. Switzerland: Springer International Publishing AG, 2019, pp. 371–380. ISBN: 0302-9743. [16] A. Myronenko. “3D MRI Brain Tumor Segmentation Using Autoencoder Regularization”. English. In: Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries. Cham: Springer International Publishing, 2019, pp. 311–320. ISBN: 0302-9743. [17] F. Isensee et al. “nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation”. In: Nature Methods 18.2 (Dec. 2020), pp. 203–211. [18] A. Hatamizadeh et al. “Swin UNETR: Swin Transformers for Semantic Segmentation of Brain Tumors in MRI Images”. English. In: (2022). [19] M. J. Cardoso et al. MONAI: An open-source framework for deep learning in healthcare. 2022. arXiv: 2211. 02701 [cs.LG]. [20] S. Ahamed et al. “Towards enhanced lesion segmentation using a 3D neural network trained on multiresolution cropped patches of lymphoma PET images”. In: Journal of Nuclear Medicine 64.supplement 1 (2023), P1360–P1360. ISSN: 0161-5505. [21] J. L. Fleiss. “Measuring nominal scale agreement among many raters”. English. In: Psychological bulletin 76.5 (1971), pp. 378–382. [22] A. K. Jha et al. “Nuclear Medicine and artificial intelligence: Best practices for evaluation (the RELAINCE guidelines)”. en. In: J. Nucl. Med. 63.9 (Sept. 2022), pp. 1288–1299. [23] N. Hasani et al. “Artificial Intelligence in Lymphoma PET Imaging: A Scoping Review (Current Trends and Future Directions)”. In: PET clinics 17.1 (Jan. 1), pp. 145–174. ISSN: 1556-8598. [24] S. K. Warfield et al. “Simultaneous truth and performance level estimation (STAPLE): an algorithm for the validation of image segmentation”. en. In: IEEE Trans. Med. Imaging 23.7 (July 2004), pp. 903–921. This paper is available on arxiv under CC 4.0 license. This paper is available on arxiv under CC 4.0 license. available on arxiv available on arxiv

Deep Neural Networks to Detect and Quantify Lymphoma Lesions: Conclusion and References

ডিপ নিউরাল নেটওয়ার্ক লিম্ফোমা ক্ষত সনাক্ত এবং পরিমাপ করার জন্য: উপসংহার এবং রেফারেন্স

Tiefe neuronale Netzwerke zur Erkennung und Quantifizierung von Lymphomläsionen: Schlussfolgerung und Referenzen

Redes neuronales profundas para detectar y cuantificar lesiones de linfoma: conclusiones y referencias

Réseaux de neurones profonds pour détecter et quantifier les lésions de lymphome : conclusion et références

लिम्फोमा घावों का पता लगाने और मात्रा निर्धारित करने के लिए डीप न्यूरल नेटवर्क: निष्कर्ष और संदर्भ

リンパ腫病変の検出と定量化のためのディープニューラルネットワーク：結論と参考文献

림프종 병변을 검출하고 정량화하기 위한 심층 신경망: 결론 및 참고 문헌

Redes Neurais Profundas para Detectar e Quantificar Lesões de Linfoma: Conclusão e Referências

Глубокие нейронные сети для обнаружения и количественной оценки поражений лимфомой: заключение и ссылки

Lenfoma Lezyonlarını Tespit Etmek ve Ölçmek İçin Derin Sinir Ağları: Sonuç ve Referanslar

Mạng lưới thần kinh sâu để phát hiện và định lượng tổn thương ung thư hạch: Kết luận và tài liệu tham khảo

深度神经网络检测和量化淋巴瘤病变：结论和参考文献

Authors:
(1) Shadab Ahamed, University of British Columbia, Vancouver, BC, Canada, BC Cancer Research Institute, Vancouver, BC, Canada. He was also a Mitacs Accelerate Fellow (May 2022 - April 2023) with Microsoft AI for Good Lab, Redmond, WA, USA (e-mail: shadabahamed1996@gmail.com);
(2) Yixi Xu, Microsoft AI for Good Lab, Redmond, WA, USA;
(3) Claire Gowdy, BC Children’s Hospital, Vancouver, BC, Canada;
(4) Joo H. O, St. Mary’s Hospital, Seoul, Republic of Korea;
(5) Ingrid Bloise, BC Cancer, Vancouver, BC, Canada;
(6) Don Wilson, BC Cancer, Vancouver, BC, Canada;
(7) Patrick Martineau, BC Cancer, Vancouver, BC, Canada;
(8) Franc¸ois Benard, BC Cancer, Vancouver, BC, Canada;
(9) Fereshteh Yousefirizi, BC Cancer Research Institute, Vancouver, BC, Canada;
(10) Rahul Dodhia, Microsoft AI for Good Lab, Redmond, WA, USA;
(11) Juan M. Lavista, Microsoft AI for Good Lab, Redmond, WA, USA;
(12) William B. Weeks, Microsoft AI for Good Lab, Redmond, WA, USA;
(13) Carlos F. Uribe, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada;
(14) Arman Rahmim, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada. Table of Links Abstract and Intro
Related Work
Materials and Methods
Results
Discussion
Conclusion and References II. RELATED WORK Numerous works have explored the application of deep learning methods for segmenting lymphoma in PET/CT images. Yuan et al. [4] developed a feature fusion technique to utilize the complementary information from multi-modality data. Hu et al. [5] proposed fusing a combination of 3D ResUNet trained on volumetric data and three 2D ResUNet trained on 2D slices from three orthogonal directions for enhancing segmentation performance. Li et al. [6] proposed DenseX-Net trained in an end-to-end fashion integrating supervised and unsupervised methods for lymphoma detection and segmentation. Liu et al. [7] introduced techniques such as patch-based negative sample augmentation and label guidance for training a 3D Residual-UNet for lymphoma segmentation. A major limitation of all these works was that they were developed on relatively smaller-sized datasets (less than 100 images). Moreover, most of these methods did not compare the performance of their proposed methods with other baselines or with the performance of physicians. Constantino et al. [8] compared the performances of 7 semi-automated and 2 deep learning segmentation methods, while Weisman et al. [9] compared 11 automated segmentation techniques, although both these studies were performed on smaller datasets of sizes 65 and 90 respectively. Weisman et al. [10] compared the segmentation performances of automated 3D Deep Medic method with that of physician although even this study included just 90 lymphoma cases. Except for [10], none of these studies reported model generalization on out-ofdistribution dataset (such as on data collected from different centers), limiting their robustness quantification and external validity. Jiang et al. [11] used a relatively larger dataset as compared to the above studies with 297 images to train a 3D UNet. They even performed out-of-distribution testing on 117 images collected from a different center. To the best of our knowledge, the largest lymphoma PET/CT dataset for deep learning-based lesion segmentation ever reported is the work by Blanc-Durand et al. [12] who used 639 images for model development and 94 for external testing; however, this study only used standard segmentation evaluation metrics and assessed their model’s ability for predicting accurate TMTV. Both the studies [11] and [12] are limited by the fact that their datasets exclusively consisted of patients diagnosed with diffuse large B-cell lymphoma (DLBCL), representing only a single subtype of lymphoma. Most of the existing studies on deep learning-based lymphoma segmentation report their performances on generic segmentation metrics such as Dice similarity coefficient (DSC), intersection-over-union (IoU), sensitivity, etc. In the presence of large segmented lesions, very small missed lesions or small false positives do not contribute much to the DSC value. Hence, there is a need to report the volumes of false positives and false negatives. It will also be beneficial to evaluate the detection performances on a per-lesion basis (number of connected components detected vs missed), since automated detection of even a few voxels of all lesions can help physicians quickly locate the regions of interest, even if the DSC is low. Moreover, the difficulty of the segmentation/detection task is often not assessed via inter- or intra-observer agreement analysis. Our study aims to address these limitations. We trained and validated four deep neural networks on lymphoma PET/CT datasets from three cohorts, encompassing two distinct subtypes of lymphoma: DLBCL and primary mediastinal large B-cell lymphoma (PMBCL). (i) We performed both in (images coming from same cohorts as the training/validation set) and out-of-distribution or external (images from a fourth cohort not used for training/validation) testing to evaluate the robustness of our models. (ii) We reported the performance using DSC, volumes of false positives and negatives, and evaluated the performance dependence on six different types of lesion measures. (iii) We also evaluated the ability of our networks to reproduce these ground truth lesion measures and computed networks’ error in predicting them. (iv) We proposed three types of detection criteria for our use-case and evaluate model’s performance on these metrics. (v) Finally, we evaluated the intra- and inter-observer agreement to give a measure of the difficulty of lesion segmentation task on our datasets. This paper is available on arxiv under CC 4.0 license. Authors: (1) Shadab Ahamed, University of British Columbia, Vancouver, BC, Canada, BC Cancer Research Institute, Vancouver, BC, Canada. He was also a Mitacs Accelerate Fellow (May 2022 - April 2023) with Microsoft AI for Good Lab, Redmond, WA, USA (e-mail: shadabahamed1996@gmail.com); (2) Yixi Xu, Microsoft AI for Good Lab, Redmond, WA, USA; (3) Claire Gowdy, BC Children’s Hospital, Vancouver, BC, Canada; (4) Joo H. O, St. Mary’s Hospital, Seoul, Republic of Korea; (5) Ingrid Bloise, BC Cancer, Vancouver, BC, Canada; (6) Don Wilson, BC Cancer, Vancouver, BC, Canada; (7) Patrick Martineau, BC Cancer, Vancouver, BC, Canada; (8) Franc¸ois Benard, BC Cancer, Vancouver, BC, Canada; (9) Fereshteh Yousefirizi, BC Cancer Research Institute, Vancouver, BC, Canada; (10) Rahul Dodhia, Microsoft AI for Good Lab, Redmond, WA, USA; (11) Juan M. Lavista, Microsoft AI for Good Lab, Redmond, WA, USA; (12) William B. Weeks, Microsoft AI for Good Lab, Redmond, WA, USA; (13) Carlos F. Uribe, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada; (14) Arman Rahmim, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada. Authors: Authors: (1) Shadab Ahamed, University of British Columbia, Vancouver, BC, Canada, BC Cancer Research Institute, Vancouver, BC, Canada. He was also a Mitacs Accelerate Fellow (May 2022 - April 2023) with Microsoft AI for Good Lab, Redmond, WA, USA (e-mail: shadabahamed1996@gmail.com); (2) Yixi Xu, Microsoft AI for Good Lab, Redmond, WA, USA; (3) Claire Gowdy, BC Children’s Hospital, Vancouver, BC, Canada; (4) Joo H. O, St. Mary’s Hospital, Seoul, Republic of Korea; (5) Ingrid Bloise, BC Cancer, Vancouver, BC, Canada; (6) Don Wilson, BC Cancer, Vancouver, BC, Canada; (7) Patrick Martineau, BC Cancer, Vancouver, BC, Canada; (8) Franc¸ois Benard, BC Cancer, Vancouver, BC, Canada; (9) Fereshteh Yousefirizi, BC Cancer Research Institute, Vancouver, BC, Canada; (10) Rahul Dodhia, Microsoft AI for Good Lab, Redmond, WA, USA; (11) Juan M. Lavista, Microsoft AI for Good Lab, Redmond, WA, USA; (12) William B. Weeks, Microsoft AI for Good Lab, Redmond, WA, USA; (13) Carlos F. Uribe, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada; (14) Arman Rahmim, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada. Table of Links Abstract and Intro Related Work Materials and Methods Results Discussion Conclusion and References Abstract and Intro Abstract and Intro Related Work Related Work Materials and Methods Materials and Methods Results Results Discussion Discussion Conclusion and References Conclusion and References II. RELATED WORK Numerous works have explored the application of deep learning methods for segmenting lymphoma in PET/CT images. Yuan et al. [4] developed a feature fusion technique to utilize the complementary information from multi-modality data. Hu et al. [5] proposed fusing a combination of 3D ResUNet trained on volumetric data and three 2D ResUNet trained on 2D slices from three orthogonal directions for enhancing segmentation performance. Li et al. [6] proposed DenseX-Net trained in an end-to-end fashion integrating supervised and unsupervised methods for lymphoma detection and segmentation. Liu et al. [7] introduced techniques such as patch-based negative sample augmentation and label guidance for training a 3D Residual-UNet for lymphoma segmentation. A major limitation of all these works was that they were developed on relatively smaller-sized datasets (less than 100 images). Moreover, most of these methods did not compare the performance of their proposed methods with other baselines or with the performance of physicians. Constantino et al. [8] compared the performances of 7 semi-automated and 2 deep learning segmentation methods, while Weisman et al. [9] compared 11 automated segmentation techniques, although both these studies were performed on smaller datasets of sizes 65 and 90 respectively. Weisman et al. [10] compared the segmentation performances of automated 3D Deep Medic method with that of physician although even this study included just 90 lymphoma cases. Except for [10], none of these studies reported model generalization on out-ofdistribution dataset (such as on data collected from different centers), limiting their robustness quantification and external validity. Jiang et al. [11] used a relatively larger dataset as compared to the above studies with 297 images to train a 3D UNet. They even performed out-of-distribution testing on 117 images collected from a different center. To the best of our knowledge, the largest lymphoma PET/CT dataset for deep learning-based lesion segmentation ever reported is the work by Blanc-Durand et al. [12] who used 639 images for model development and 94 for external testing; however, this study only used standard segmentation evaluation metrics and assessed their model’s ability for predicting accurate TMTV. Both the studies [11] and [12] are limited by the fact that their datasets exclusively consisted of patients diagnosed with diffuse large B-cell lymphoma (DLBCL), representing only a single subtype of lymphoma. Most of the existing studies on deep learning-based lymphoma segmentation report their performances on generic segmentation metrics such as Dice similarity coefficient (DSC), intersection-over-union (IoU), sensitivity, etc. In the presence of large segmented lesions, very small missed lesions or small false positives do not contribute much to the DSC value. Hence, there is a need to report the volumes of false positives and false negatives. It will also be beneficial to evaluate the detection performances on a per-lesion basis (number of connected components detected vs missed), since automated detection of even a few voxels of all lesions can help physicians quickly locate the regions of interest, even if the DSC is low. Moreover, the difficulty of the segmentation/detection task is often not assessed via inter- or intra-observer agreement analysis. Our study aims to address these limitations. We trained and validated four deep neural networks on lymphoma PET/CT datasets from three cohorts, encompassing two distinct subtypes of lymphoma: DLBCL and primary mediastinal large B-cell lymphoma (PMBCL). (i) We performed both in (images coming from same cohorts as the training/validation set) and out-of-distribution or external (images from a fourth cohort not used for training/validation) testing to evaluate the robustness of our models. (ii) We reported the performance using DSC, volumes of false positives and negatives, and evaluated the performance dependence on six different types of lesion measures. (iii) We also evaluated the ability of our networks to reproduce these ground truth lesion measures and computed networks’ error in predicting them. (iv) We proposed three types of detection criteria for our use-case and evaluate model’s performance on these metrics. (v) Finally, we evaluated the intra- and inter-observer agreement to give a measure of the difficulty of lesion segmentation task on our datasets. This paper is available on arxiv under CC 4.0 license. This paper is available on arxiv under CC 4.0 license. available on arxiv

Deep Neural Networks to Detect and Quantify Lymphoma Lesions: Related Work

ডিপ নিউরাল নেটওয়ার্ক লিম্ফোমা ক্ষত সনাক্ত এবং পরিমাপ করতে: সম্পর্কিত কাজ

Tiefe neuronale Netzwerke zur Erkennung und Quantifizierung von Lymphomläsionen: Verwandte Arbeiten

Redes neuronales profundas para detectar y cuantificar lesiones de linfoma: trabajo relacionado

Réseaux de neurones profonds pour détecter et quantifier les lésions de lymphome : travaux connexes

लिम्फोमा घावों का पता लगाने और उनकी मात्रा निर्धारित करने के लिए डीप न्यूरल नेटवर्क: संबंधित कार्य

リンパ腫病変の検出と定量化のためのディープニューラルネットワーク: 関連研究

림프종 병변을 탐지하고 정량화하기 위한 심층 신경망: 관련 연구

Redes Neurais Profundas para Detectar e Quantificar Lesões de Linfoma: Trabalho Relacionado

Глубокие нейронные сети для обнаружения и количественной оценки поражений лимфомой: соответствующая работа

Lenfoma Lezyonlarını Tespit ve Ölçmek İçin Derin Sinir Ağları: İlgili Çalışma

Mạng lưới thần kinh sâu để phát hiện và định lượng tổn thương ung thư hạch: Công việc liên quan

深度神经网络检测和量化淋巴瘤病变：相关工作

Authors:
(1) Shadab Ahamed, University of British Columbia, Vancouver, BC, Canada, BC Cancer Research Institute, Vancouver, BC, Canada. He was also a Mitacs Accelerate Fellow (May 2022 - April 2023) with Microsoft AI for Good Lab, Redmond, WA, USA (e-mail: shadabahamed1996@gmail.com);
(2) Yixi Xu, Microsoft AI for Good Lab, Redmond, WA, USA;
(3) Claire Gowdy, BC Children’s Hospital, Vancouver, BC, Canada;
(4) Joo H. O, St. Mary’s Hospital, Seoul, Republic of Korea;
(5) Ingrid Bloise, BC Cancer, Vancouver, BC, Canada;
(6) Don Wilson, BC Cancer, Vancouver, BC, Canada;
(7) Patrick Martineau, BC Cancer, Vancouver, BC, Canada;
(8) Franc¸ois Benard, BC Cancer, Vancouver, BC, Canada;
(9) Fereshteh Yousefirizi, BC Cancer Research Institute, Vancouver, BC, Canada;
(10) Rahul Dodhia, Microsoft AI for Good Lab, Redmond, WA, USA;
(11) Juan M. Lavista, Microsoft AI for Good Lab, Redmond, WA, USA;
(12) William B. Weeks, Microsoft AI for Good Lab, Redmond, WA, USA;
(13) Carlos F. Uribe, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada;
(14) Arman Rahmim, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada. Table of Links Abstract and Intro
Related Work
Materials and Methods
Results
Discussion
Conclusion and References V. DISCUSSION In this work, we trained and evaluated four distinct neural network architectures to automate the segmentation of lymphoma lesions from PET/CT datasets sourced from three different cohorts. To assess models performance, we conducted comprehensive evaluations on internal test set originating from these three cohorts and showed that SegResNet and UNet outperformed DynUNet and SwinUNETR on the DSC (mean and median) and median FPV metrics, while SwinUNETR had the best median FNV. In addition to internal evaluations, we extended our analysis to encompass an external outof-distribution testing phase on a sizable public lymphoma PET/CT dataset. On this external test set as well, SegResNet emerged as the top performer in terms of DSC and FPV metrics, underscoring its robustness and effectiveness, while UNet displayed the best performance on FNV. It is important to highlight that SegResNet and UNet were trained on patches of larger sizes, specifically (224, 224, 224) and (192, 192, 192) respectively, while DynUNet and SwinUNETR were trained using relatively smaller patches, namely (160, 160, 160) and (128, 128, 128) respectively. Utilizing larger patch sizes during training allows the neural networks to capture a more extensive contextual understanding of the data, thereby enhancing its performance in segmentation tasks [17]. This observation aligns with our results, where the superior performance of SegResNet and UNet can be attributed to their exposure to larger patch sizes during training. Moreover, larger batch sizes enable robust training by accurately estimating the gradients [17], but with our chosen training patch sizes, we could not train SegResNet, DynUNet and SwinUNETR with nb > 1 due to memory limitations (although we could accommodate nb = 8 for UNet). Hence, for a fair comparison between networks, all networks were trained with nb = 1. It is worth noting that our inability to train DynUNet and SwinUNETR on larger patch and mini-batch sizes was primarily due to computational resource limitations. However, this limitation presents an avenue for future research, where training these models with larger patches and batch sizes could potentially yield further improvements in segmentation accuracy. We assessed the reproducibility of lesions measures and found that on the internal test set, TMTV and TLG were reproducible across all networks, while Dmax was not reproducible by any network. SUVmean was reproducible by all networks except UNet, SUVmax by only SegResNet and number of lesions by only UNet and SegResNet. On the external test set, reproducibility was more limited, with only SUVmean being reproducible by both SegResNet and SwinUNETR, number of lesions by SegResNet, and TLG by DynUNet (Fig. 3 and 4). Furthermore, we quantified the networks’ error in estimating the value of lesion measures using MAPE and found that MAPE generally decreases as a function of lesion measure values (for all lesion measures) on the combined internal and external test set (Fig. 5). The networks generally made significant errors in the accurate prediction when the ground truth lesion measures were very small. We also showed that, in general, on a set of images with larger patient level lesion SUVmean, SUVmean, TMTV, and TLG, a network is able to predict a higher median DSC, although for very high values of these lesion measures, the performance generally plateaus. On the other hand, the DSC performance is not much affected by the number of lesions, while for a set of images with higher Dmax, the performance generally decreases for all networks (Fig. 7). As much of PET/CT data is privately owned by healthcare institutions, it poses significant challenges for researchers in accessing diverse datasets for training and testing deep learning models. In such a scenario, to improve the interpretability of models, it is crucial for researchers to investigate how the performance of their models depend on dataset characteristics. By studying how model performance correlates with the image/lesion characteristics, researchers can gain insights into the strengths and limitations of their models [13]. Alongside the evaluation of segmentation performance, we also introduced three distinct detection criteria, denoted as Criterion 1, 2, and 3. These criteria served a specific purpose: to evaluate the networks’ performance on a per-lesion basis. This stands in contrast to the segmentation performance assessment, which primarily focuses on the voxel-level accuracy of the networks. The rationale behind introducing these detection criteria lies in the need to assess how well the networks identify and detect lesions within the images, as opposed to merely evaluating their ability to delineate lesion boundaries at the voxel level. The ability to detect the presence of lesions (Criterion 1) is crucial, as it directly influences whether a potential health concern is identified or missed. Detecting even a single voxel of a lesion could trigger further investigation or treatment planning. Lesion count and accurate localization (Criterion 2) are important for treatment planning and monitoring disease progression. Knowing not only that a lesion exists but also how many there are and where they are located can significantly impact therapeutic decisions. Criterion 3 which focused on segmenting lesions based on lesion metabolic characteristics (SUVmax), adds an additional layer of clinical relevance. Using these detection metrics, we assessed the sensitivities and FP detections for all networks and showed that depending on the detection criteria, a network can have very high sensitivity even when the DSC performance was low. Given these different detection criteria, a trained model can be chosen based on specific clinical use cases. For example, some use cases might involve being able to detect all lesions without being overly cautious about segmenting exact lesion boundary, while some other use cases might be looking for more robust boundary delineations. Furthermore, we assessed the intra-observer variability of a physician in segmenting both “easy” and “hard” cases, noting challenges in consistent segmentation of cases from the “hard” subset. In lymphoma lesion segmentation, cases can vary in difficulty due to factors like size, shape, and location of lesions, or image quality. By identifying which cases are consistently difficult for even an experienced physician to segment, we gained insights into the complexities and nuances of the segmentation task. Finally, we also assessed the interobserver agreement between three physicians. Although, we inferred that there was substantial level of agreement between the three physicians, the assessment was performed only on 9 cases, resulting in low statistical power. To improve the consistency of ground truth in medical image segmentation, a well-defined protocol is essential. This protocol should engage multiple expert physicians independently in delineating regions of interest (ROIs) or lesions within PET/CT images. Instead of a single physician segmenting a cohort independently, multiple annotators should segment the same images without knowledge of each other’s work. Discrepancies or disagreements among physicians can be resolved through structured approaches such as facilitated discussions, clinical information reviews, or image clarification. This robust ground truth process enhances inter-observer agreement accuracy and strengthens the validity of research findings and clinical applications relying on these annotations. This paper is available on arxiv under CC 4.0 license. Authors: (1) Shadab Ahamed, University of British Columbia, Vancouver, BC, Canada, BC Cancer Research Institute, Vancouver, BC, Canada. He was also a Mitacs Accelerate Fellow (May 2022 - April 2023) with Microsoft AI for Good Lab, Redmond, WA, USA (e-mail: shadabahamed1996@gmail.com); (2) Yixi Xu, Microsoft AI for Good Lab, Redmond, WA, USA; (3) Claire Gowdy, BC Children’s Hospital, Vancouver, BC, Canada; (4) Joo H. O, St. Mary’s Hospital, Seoul, Republic of Korea; (5) Ingrid Bloise, BC Cancer, Vancouver, BC, Canada; (6) Don Wilson, BC Cancer, Vancouver, BC, Canada; (7) Patrick Martineau, BC Cancer, Vancouver, BC, Canada; (8) Franc¸ois Benard, BC Cancer, Vancouver, BC, Canada; (9) Fereshteh Yousefirizi, BC Cancer Research Institute, Vancouver, BC, Canada; (10) Rahul Dodhia, Microsoft AI for Good Lab, Redmond, WA, USA; (11) Juan M. Lavista, Microsoft AI for Good Lab, Redmond, WA, USA; (12) William B. Weeks, Microsoft AI for Good Lab, Redmond, WA, USA; (13) Carlos F. Uribe, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada; (14) Arman Rahmim, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada. Authors: Authors: (1) Shadab Ahamed, University of British Columbia, Vancouver, BC, Canada, BC Cancer Research Institute, Vancouver, BC, Canada. He was also a Mitacs Accelerate Fellow (May 2022 - April 2023) with Microsoft AI for Good Lab, Redmond, WA, USA (e-mail: shadabahamed1996@gmail.com); (2) Yixi Xu, Microsoft AI for Good Lab, Redmond, WA, USA; (3) Claire Gowdy, BC Children’s Hospital, Vancouver, BC, Canada; (4) Joo H. O, St. Mary’s Hospital, Seoul, Republic of Korea; (5) Ingrid Bloise, BC Cancer, Vancouver, BC, Canada; (6) Don Wilson, BC Cancer, Vancouver, BC, Canada; (7) Patrick Martineau, BC Cancer, Vancouver, BC, Canada; (8) Franc¸ois Benard, BC Cancer, Vancouver, BC, Canada; (9) Fereshteh Yousefirizi, BC Cancer Research Institute, Vancouver, BC, Canada; (10) Rahul Dodhia, Microsoft AI for Good Lab, Redmond, WA, USA; (11) Juan M. Lavista, Microsoft AI for Good Lab, Redmond, WA, USA; (12) William B. Weeks, Microsoft AI for Good Lab, Redmond, WA, USA; (13) Carlos F. Uribe, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada; (14) Arman Rahmim, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada. Table of Links Abstract and Intro Related Work Materials and Methods Results Discussion Conclusion and References Abstract and Intro Abstract and Intro Related Work Related Work Materials and Methods Materials and Methods Results Results Discussion Discussion Conclusion and References Conclusion and References V. DISCUSSION In this work, we trained and evaluated four distinct neural network architectures to automate the segmentation of lymphoma lesions from PET/CT datasets sourced from three different cohorts. To assess models performance, we conducted comprehensive evaluations on internal test set originating from these three cohorts and showed that SegResNet and UNet outperformed DynUNet and SwinUNETR on the DSC (mean and median) and median FPV metrics, while SwinUNETR had the best median FNV. In addition to internal evaluations, we extended our analysis to encompass an external outof-distribution testing phase on a sizable public lymphoma PET/CT dataset. On this external test set as well, SegResNet emerged as the top performer in terms of DSC and FPV metrics, underscoring its robustness and effectiveness, while UNet displayed the best performance on FNV. It is important to highlight that SegResNet and UNet were trained on patches of larger sizes, specifically (224, 224, 224) and (192, 192, 192) respectively, while DynUNet and SwinUNETR were trained using relatively smaller patches, namely (160, 160, 160) and (128, 128, 128) respectively. Utilizing larger patch sizes during training allows the neural networks to capture a more extensive contextual understanding of the data, thereby enhancing its performance in segmentation tasks [17]. This observation aligns with our results, where the superior performance of SegResNet and UNet can be attributed to their exposure to larger patch sizes during training. Moreover, larger batch sizes enable robust training by accurately estimating the gradients [17], but with our chosen training patch sizes, we could not train SegResNet, DynUNet and SwinUNETR with nb > 1 due to memory limitations (although we could accommodate nb = 8 for UNet). Hence, for a fair comparison between networks, all networks were trained with nb = 1. It is worth noting that our inability to train DynUNet and SwinUNETR on larger patch and mini-batch sizes was primarily due to computational resource limitations. However, this limitation presents an avenue for future research, where training these models with larger patches and batch sizes could potentially yield further improvements in segmentation accuracy. We assessed the reproducibility of lesions measures and found that on the internal test set, TMTV and TLG were reproducible across all networks, while Dmax was not reproducible by any network. SUVmean was reproducible by all networks except UNet, SUVmax by only SegResNet and number of lesions by only UNet and SegResNet. On the external test set, reproducibility was more limited, with only SUVmean being reproducible by both SegResNet and SwinUNETR, number of lesions by SegResNet, and TLG by DynUNet (Fig. 3 and 4). Furthermore, we quantified the networks’ error in estimating the value of lesion measures using MAPE and found that MAPE generally decreases as a function of lesion measure values (for all lesion measures) on the combined internal and external test set (Fig. 5). The networks generally made significant errors in the accurate prediction when the ground truth lesion measures were very small. We also showed that, in general, on a set of images with larger patient level lesion SUVmean, SUVmean, TMTV, and TLG, a network is able to predict a higher median DSC, although for very high values of these lesion measures, the performance generally plateaus. On the other hand, the DSC performance is not much affected by the number of lesions, while for a set of images with higher Dmax, the performance generally decreases for all networks (Fig. 7). As much of PET/CT data is privately owned by healthcare institutions, it poses significant challenges for researchers in accessing diverse datasets for training and testing deep learning models. In such a scenario, to improve the interpretability of models, it is crucial for researchers to investigate how the performance of their models depend on dataset characteristics. By studying how model performance correlates with the image/lesion characteristics, researchers can gain insights into the strengths and limitations of their models [13]. Alongside the evaluation of segmentation performance, we also introduced three distinct detection criteria, denoted as Criterion 1, 2, and 3. These criteria served a specific purpose: to evaluate the networks’ performance on a per-lesion basis. This stands in contrast to the segmentation performance assessment, which primarily focuses on the voxel-level accuracy of the networks. The rationale behind introducing these detection criteria lies in the need to assess how well the networks identify and detect lesions within the images, as opposed to merely evaluating their ability to delineate lesion boundaries at the voxel level. The ability to detect the presence of lesions (Criterion 1) is crucial, as it directly influences whether a potential health concern is identified or missed. Detecting even a single voxel of a lesion could trigger further investigation or treatment planning. Lesion count and accurate localization (Criterion 2) are important for treatment planning and monitoring disease progression. Knowing not only that a lesion exists but also how many there are and where they are located can significantly impact therapeutic decisions. Criterion 3 which focused on segmenting lesions based on lesion metabolic characteristics (SUVmax), adds an additional layer of clinical relevance. Using these detection metrics, we assessed the sensitivities and FP detections for all networks and showed that depending on the detection criteria, a network can have very high sensitivity even when the DSC performance was low. Given these different detection criteria, a trained model can be chosen based on specific clinical use cases. For example, some use cases might involve being able to detect all lesions without being overly cautious about segmenting exact lesion boundary, while some other use cases might be looking for more robust boundary delineations. Furthermore, we assessed the intra-observer variability of a physician in segmenting both “easy” and “hard” cases, noting challenges in consistent segmentation of cases from the “hard” subset. In lymphoma lesion segmentation, cases can vary in difficulty due to factors like size, shape, and location of lesions, or image quality. By identifying which cases are consistently difficult for even an experienced physician to segment, we gained insights into the complexities and nuances of the segmentation task. Finally, we also assessed the interobserver agreement between three physicians. Although, we inferred that there was substantial level of agreement between the three physicians, the assessment was performed only on 9 cases, resulting in low statistical power. To improve the consistency of ground truth in medical image segmentation, a well-defined protocol is essential. This protocol should engage multiple expert physicians independently in delineating regions of interest (ROIs) or lesions within PET/CT images. Instead of a single physician segmenting a cohort independently, multiple annotators should segment the same images without knowledge of each other’s work. Discrepancies or disagreements among physicians can be resolved through structured approaches such as facilitated discussions, clinical information reviews, or image clarification. This robust ground truth process enhances inter-observer agreement accuracy and strengthens the validity of research findings and clinical applications relying on these annotations. This paper is available on arxiv under CC 4.0 license. This paper is available on arxiv under CC 4.0 license. available on arxiv available on arxiv

Deep Neural Networks to Detect and Quantify Lymphoma Lesions: Discussion

লিম্ফোমা ক্ষত সনাক্ত এবং পরিমাপ করার জন্য গভীর নিউরাল নেটওয়ার্ক: আলোচনা

Tiefe neuronale Netzwerke zur Erkennung und Quantifizierung von Lymphomläsionen: Diskussion

Redes neuronales profundas para detectar y cuantificar lesiones de linfoma: discusión

Réseaux de neurones profonds pour détecter et quantifier les lésions de lymphome : discussion

लिम्फोमा घावों का पता लगाने और मात्रा निर्धारित करने के लिए डीप न्यूरल नेटवर्क: चर्चा

リンパ腫病変を検出し定量化するディープニューラルネットワーク：議論

림프종 병변을 탐지하고 정량화하기 위한 심층 신경망: 토론

Redes Neurais Profundas para Detectar e Quantificar Lesões de Linfoma: Discussão

Глубокие нейронные сети для обнаружения и количественной оценки поражений лимфомой: обсуждение

Lenfoma Lezyonlarını Tespit Etmek ve Ölçmek İçin Derin Sinir Ağları: Tartışma

Mạng lưới thần kinh sâu để phát hiện và định lượng tổn thương ung thư hạch: Thảo luận

深度神经网络检测和量化淋巴瘤病变：讨论

Author:
(1) Shadab Ahamed, University of British Columbia, Vancouver, BC, Canada, BC Cancer Research Institute, Vancouver, BC, Canada. He was also a Mitacs Accelerate Fellow (May 2022 - April 2023) with Microsoft AI for Good Lab, Redmond, WA, USA (e-mail: shadabahamed1996@gmail.com);
(2) Yixi Xu, Microsoft AI for Good Lab, Redmond, WA, USA;
(3) Claire Gowdy, BC Children’s Hospital, Vancouver, BC, Canada;
(4) Joo H. O, St. Mary’s Hospital, Seoul, Republic of Korea;
(5) Ingrid Bloise, BC Cancer, Vancouver, BC, Canada;
(6) Don Wilson, BC Cancer, Vancouver, BC, Canada;
(7) Patrick Martineau, BC Cancer, Vancouver, BC, Canada;
(8) Franc¸ois Benard, BC Cancer, Vancouver, BC, Canada;
(9) Fereshteh Yousefirizi, BC Cancer Research Institute, Vancouver, BC, Canada;
(10) Rahul Dodhia, Microsoft AI for Good Lab, Redmond, WA, USA;
(11) Juan M. Lavista, Microsoft AI for Good Lab, Redmond, WA, USA;
(12) William B. Weeks, Microsoft AI for Good Lab, Redmond, WA, USA;
(13) Carlos F. Uribe, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada;
(14) Arman Rahmim, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada. Table of Links Abstract and Intro
Related Work
Materials and Methods
Results
Discussion
Conclusion and References Abstract This study performs comprehensive evaluation of four neural network architectures (UNet, SegResNet, DynUNet, and SwinUNETR) for lymphoma lesion segmentation from PET/CT images. These networks were trained, validated and tested on a diverse, multi-institutional dataset of 611 cases. Internal testing (88 cases; total metabolic tumor volume (TMTV) range [0.52, 2300] ml) showed SegResNet as the top performer with a median Dice similarity coefficient (DSC) of 0.76 and median false positive volume (FPV) of 4.55 ml; all networks had a median false negative volume (FNV) of 0 ml. On the unseen external test set (145 cases with TMTV range: [0.10, 2480] ml), SegResNet achieved the best median DSC of 0.68 and FPV of 21.46 ml, while UNet had the best FNV of 0.41 ml. We assessed reproducibility of six lesion measures, calculated their prediction errors, and examined DSC performance in relation to these lesion measures, offering insights into segmentation accuracy and clinical relevance. Additionally, we introduced three lesion detection criteria, addressing the clinical need for identifying lesions, counting them, and segmenting based on metabolic characteristics. We also performed expert intra-observer variability analysis revealing the challenges in segmenting “easy” vs. “hard” cases, to assist in the development of more resilient segmentation algorithms. Finally, we performed inter-observer agreement assessment underscoring the importance of a standardized ground truth segmentation protocol involving multiple expert annotators. Code is available at: https://github.com/microsoft/lymphoma-segmentationdnn. Index Terms— Positron emission tomography, computed tomography, deep learning, segmentation, detection, lesion measures, intra-observer variability, inter-observer variability I. INTRODUCTION F LUORODEOXYGLUCOSE (18F-FDG) PET/CT imaging is the standard of care for lymphoma patients, providing accurate diagnoses, staging, and therapy response evaluation. However, traditional qualitative assessments, like Deauville scores [1], can introduce variability due to observer’s subjectivity in image interpretation. Using quantitative PET analysis that incorporates lesion measures such as mean lesion standardized uptake value (SUVmean), total metabolic tumor volume (TMTV), and total lesion glycolysis (TLG) offers a promising path to more reliable prognostic decisions, enhancing our ability to predict patient outcomes in lymphoma with greater precision and confidence [2]. Quantitative assessment in PET/CT imaging often relies on manual lesion segmentation, which is time-consuming and prone to intra- and inter-observer variabilities. Traditional thresholding-based automated techniques can miss low-uptake disease and produce false positives in regions of physiological high uptake of radiotracers. Therefore, deep learning offers promise for automating lesion segmentation, reducing variability, increasing patient throughput, and potentially aiding in the detection of challenging lesions [3]. Although promising, deep learning methods face challenges of their own. Convolutional neural networks (CNNs) require large, well-annotated datasets that can be difficult to obtain. Models trained on small datasets may not be generalizable. Moreover, lymphoma lesions vary significantly in size, shape, and metabolic activity, making training deep networks accurately challenging in the absence of well-defined priors. Deep learning aims to reduce observer variability, but inconsistent manual annotations used for training can lead to error perpetuation. Understanding these challenges is crucial towards harnessing the full potential of these methods in PET/CT quantitative analysis. This paper is available on arxiv under CC 4.0 license. Author: (1) Shadab Ahamed, University of British Columbia, Vancouver, BC, Canada, BC Cancer Research Institute, Vancouver, BC, Canada. He was also a Mitacs Accelerate Fellow (May 2022 - April 2023) with Microsoft AI for Good Lab, Redmond, WA, USA (e-mail: shadabahamed1996@gmail.com); (2) Yixi Xu, Microsoft AI for Good Lab, Redmond, WA, USA; (3) Claire Gowdy, BC Children’s Hospital, Vancouver, BC, Canada; (4) Joo H. O, St. Mary’s Hospital, Seoul, Republic of Korea; (5) Ingrid Bloise, BC Cancer, Vancouver, BC, Canada; (6) Don Wilson, BC Cancer, Vancouver, BC, Canada; (7) Patrick Martineau, BC Cancer, Vancouver, BC, Canada; (8) Franc¸ois Benard, BC Cancer, Vancouver, BC, Canada; (9) Fereshteh Yousefirizi, BC Cancer Research Institute, Vancouver, BC, Canada; (10) Rahul Dodhia, Microsoft AI for Good Lab, Redmond, WA, USA; (11) Juan M. Lavista, Microsoft AI for Good Lab, Redmond, WA, USA; (12) William B. Weeks, Microsoft AI for Good Lab, Redmond, WA, USA; (13) Carlos F. Uribe, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada; (14) Arman Rahmim, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada. Author: Author: (1) Shadab Ahamed, University of British Columbia, Vancouver, BC, Canada, BC Cancer Research Institute, Vancouver, BC, Canada. He was also a Mitacs Accelerate Fellow (May 2022 - April 2023) with Microsoft AI for Good Lab, Redmond, WA, USA (e-mail: shadabahamed1996@gmail.com); (2) Yixi Xu, Microsoft AI for Good Lab, Redmond, WA, USA; (3) Claire Gowdy, BC Children’s Hospital, Vancouver, BC, Canada; (4) Joo H. O, St. Mary’s Hospital, Seoul, Republic of Korea; (5) Ingrid Bloise, BC Cancer, Vancouver, BC, Canada; (6) Don Wilson, BC Cancer, Vancouver, BC, Canada; (7) Patrick Martineau, BC Cancer, Vancouver, BC, Canada; (8) Franc¸ois Benard, BC Cancer, Vancouver, BC, Canada; (9) Fereshteh Yousefirizi, BC Cancer Research Institute, Vancouver, BC, Canada; (10) Rahul Dodhia, Microsoft AI for Good Lab, Redmond, WA, USA; (11) Juan M. Lavista, Microsoft AI for Good Lab, Redmond, WA, USA; (12) William B. Weeks, Microsoft AI for Good Lab, Redmond, WA, USA; (13) Carlos F. Uribe, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada; (14) Arman Rahmim, BC Cancer Research Institute, Vancouver, BC, Canada, and University of British Columbia, Vancouver, BC, Canada. Table of Links Abstract and Intro Related Work Materials and Methods Results Discussion Conclusion and References Abstract and Intro Abstract and Intro Related Work Related Work Materials and Methods Materials and Methods Results Results Discussion Discussion Conclusion and References Conclusion and References Abstract Abstract This study performs comprehensive evaluation of four neural network architectures (UNet, SegResNet, DynUNet, and SwinUNETR) for lymphoma lesion segmentation from PET/CT images. These networks were trained, validated and tested on a diverse, multi-institutional dataset of 611 cases. Internal testing (88 cases; total metabolic tumor volume (TMTV) range [0.52, 2300] ml) showed SegResNet as the top performer with a median Dice similarity coefficient (DSC) of 0.76 and median false positive volume (FPV) of 4.55 ml; all networks had a median false negative volume (FNV) of 0 ml. On the unseen external test set (145 cases with TMTV range: [0.10, 2480] ml), SegResNet achieved the best median DSC of 0.68 and FPV of 21.46 ml, while UNet had the best FNV of 0.41 ml. We assessed reproducibility of six lesion measures, calculated their prediction errors, and examined DSC performance in relation to these lesion measures, offering insights into segmentation accuracy and clinical relevance. Additionally, we introduced three lesion detection criteria, addressing the clinical need for identifying lesions, counting them, and segmenting based on metabolic characteristics. We also performed expert intra-observer variability analysis revealing the challenges in segmenting “easy” vs. “hard” cases, to assist in the development of more resilient segmentation algorithms. Finally, we performed inter-observer agreement assessment underscoring the importance of a standardized ground truth segmentation protocol involving multiple expert annotators. Code is available at: https://github.com/microsoft/lymphoma-segmentationdnn . https://github.com/microsoft/lymphoma-segmentationdnn Index Terms — Positron emission tomography, computed tomography, deep learning, segmentation, detection, lesion measures, intra-observer variability, inter-observer variability Index Terms I. INTRODUCTION F LUORODEOXYGLUCOSE (18F-FDG) PET/CT imaging is the standard of care for lymphoma patients, providing accurate diagnoses, staging, and therapy response evaluation. However, traditional qualitative assessments, like Deauville scores [1], can introduce variability due to observer’s subjectivity in image interpretation. Using quantitative PET analysis that incorporates lesion measures such as mean lesion standardized uptake value (SUVmean), total metabolic tumor volume (TMTV), and total lesion glycolysis (TLG) offers a promising path to more reliable prognostic decisions, enhancing our ability to predict patient outcomes in lymphoma with greater precision and confidence [2]. Quantitative assessment in PET/CT imaging often relies on manual lesion segmentation, which is time-consuming and prone to intra- and inter-observer variabilities. Traditional thresholding-based automated techniques can miss low-uptake disease and produce false positives in regions of physiological high uptake of radiotracers. Therefore, deep learning offers promise for automating lesion segmentation, reducing variability, increasing patient throughput, and potentially aiding in the detection of challenging lesions [3]. Although promising, deep learning methods face challenges of their own. Convolutional neural networks (CNNs) require large, well-annotated datasets that can be difficult to obtain. Models trained on small datasets may not be generalizable. Moreover, lymphoma lesions vary significantly in size, shape, and metabolic activity, making training deep networks accurately challenging in the absence of well-defined priors. Deep learning aims to reduce observer variability, but inconsistent manual annotations used for training can lead to error perpetuation. Understanding these challenges is crucial towards harnessing the full potential of these methods in PET/CT quantitative analysis. This paper is available on arxiv under CC 4.0 license. This paper is available on arxiv under CC 4.0 license. available on arxiv

Deep Neural Networks to Detect and Quantify Lymphoma Lesions: Abstract and Intro

লিম্ফোমা ক্ষত সনাক্ত এবং পরিমাপ করার জন্য গভীর নিউরাল নেটওয়ার্ক: বিমূর্ত এবং ভূমিকা

Tiefe neuronale Netzwerke zur Erkennung und Quantifizierung von Lymphomläsionen: Zusammenfassung und Einführung

Redes neuronales profundas para detectar y cuantificar lesiones de linfoma: resumen e introducción

Réseaux de neurones profonds pour détecter et quantifier les lésions de lymphome : résumé et introduction

लिम्फोमा घावों का पता लगाने और मात्रा निर्धारित करने के लिए डीप न्यूरल नेटवर्क: सार और परिचय

リンパ腫病変を検出し定量化するディープニューラルネットワーク：概要と概要

림프종 병변을 탐지하고 정량화하기 위한 심층 신경망: 개요 및 소개

Redes Neurais Profundas para Detectar e Quantificar Lesões de Linfoma: Resumo e Introdução

Глубокие нейронные сети для обнаружения и количественной оценки поражений лимфомой: краткое содержание и введение

Lenfoma Lezyonlarını Tespit Etmek ve Ölçmek İçin Derin Sinir Ağları: Özet ve Giriş

Mạng lưới thần kinh sâu để phát hiện và định lượng các tổn thương ung thư hạch: Tóm tắt và giới thiệu

深度神经网络检测和量化淋巴瘤病变：摘要和简介

In the land down under, there’s a tale that’s taken flight, literally. Picture this: one lone soul in Perth, Western Australia, is shaking up the skies with complaints about aircraft noise. Move over, kangaroos boxing on the street; this is the real drama unfolding! Now, get ready to have your socks knocked off by numbers. Out of over 51,000 aircraft noise complaints across Australia, nearly 21,000 of them were dropped by our Perth protagonist. That’s some serious air rage! But wait, it gets better. Brisbane tried to join the complaint party, but they only managed to squeak out around 4,700 gripes. Sorry, Brisbane, but Perth is owning the noise game! Now, before you start laughing it off as a case of someone just having a rough day, think again. Studies show that constant aircraft noise can turn even the most chilled Aussie into a grumpy wallaby. We’re talking sleep disturbances, stress levels skyrocketing higher than a kangaroo on a pogo stick—real chaos, mate! And it’s not just about personal comfort. Oh no, we’re diving into the big bucks territory. Property values? Plummeting faster than a boomerang thrown by a toddler! Now, here’s where it gets interesting. We’re not just talking about a random person shouting into the wind. Nope, this Perth warrior is the David to the aviation Goliath, taking on the system with a vengeance! And speaking of systems, Airservices Australia noticed a curious trend: fewer people are making more noise complaints. It’s like everyone’s decided to let this one Aussie do all the heavy lifting, while the rest of the country enjoys their barbies and surf sessions. But let’s not forget the heart of the matter. These aren’t just complaints; they’re battle cries against the tyranny of aircraft noise. It’s like a scene straight out of "Crocodile Dundee," but instead of wrestling crocs, we’re wrestling with jet engines! And while the suits in Canberra debate policies and regulations, our Perth hero is out there, reminding us that one person’s voice can make a noise big enough to echo across the whole continent. So, next time you hear a plane overhead, spare a thought for our Perth pal fighting the good fight against the roar of the engines. And maybe, just maybe, invest in some noise-canceling headphones. So for the Joke! What did the man who lived near an international airport, one of Australia’s international hubs do when he got sick of all the aircraft noise? Answer #. He moved. In the land down under, there’s a tale that’s taken flight, literally. Picture this: one lone soul in Perth, Western Australia, is shaking up the skies with complaints about aircraft noise. Move over, kangaroos boxing on the street; this is the real drama unfolding! Now, get ready to have your socks knocked off by numbers. Out of over 51,000 aircraft noise complaints across Australia , nearly 21,000 of them were dropped by our Perth protagonist. That’s some serious air rage! aircraft noise complaints across Australia But wait, it gets better. Brisbane tried to join the complaint party, but they only managed to squeak out around 4,700 gripes. Sorry, Brisbane, but Perth is owning the noise game! Now, before you start laughing it off as a case of someone just having a rough day, think again. Studies show that constant aircraft noise can turn even the most chilled Aussie into a grumpy wallaby. We’re talking sleep disturbances, stress levels skyrocketing higher than a kangaroo on a pogo stick—real chaos, mate! And it’s not just about personal comfort. Oh no, we’re diving into the big bucks territory. Property values? Plummeting faster than a boomerang thrown by a toddler! Now, here’s where it gets interesting. We’re not just talking about a random person shouting into the wind. Nope, this Perth warrior is the David to the aviation Goliath, taking on the system with a vengeance! And speaking of systems, Airservices Australia noticed a curious trend: fewer people are making more noise complaints. It’s like everyone’s decided to let this one Aussie do all the heavy lifting, while the rest of the country enjoys their barbies and surf sessions. But let’s not forget the heart of the matter. These aren’t just complaints; they’re battle cries against the tyranny of aircraft noise. It’s like a scene straight out of "Crocodile Dundee," but instead of wrestling crocs, we’re wrestling with jet engines! And while the suits in Canberra debate policies and regulations, our Perth hero is out there, reminding us that one person’s voice can make a noise big enough to echo across the whole continent. So, next time you hear a plane overhead, spare a thought for our Perth pal fighting the good fight against the roar of the engines. And maybe, just maybe, invest in some noise-canceling headphones. So for the Joke! What did the man who lived near an international airport, one of Australia’s international hubs do when he got sick of all the aircraft noise? Answer #. He moved.

The Funny Australian Aircraft Noise Complaint  

Authors:
(1) Sasun Hambardzumyan, Activeloop, Mountain View, CA, USA;
(2) Abhinav Tuli, Activeloop, Mountain View, CA, USA;
(3) Levon Ghukasyan, Activeloop, Mountain View, CA, USA;
(4) Fariz Rahman, Activeloop, Mountain View, CA, USA;.
(5) Hrant Topchyan, Activeloop, Mountain View, CA, USA;
(6) David Isayan, Activeloop, Mountain View, CA, USA;
(7) Mark McQuade, Activeloop, Mountain View, CA, USA;
(8) Mikayel Harutyunyan, Activeloop, Mountain View, CA, USA;
(9) Tatevik Hakobyan, Activeloop, Mountain View, CA, USA;
(10) Ivo Stranic, Activeloop, Mountain View, CA, USA;
(11) Davit Buniatyan, Activeloop, Mountain View, CA, USA. Table of Links Abstract and Intro
Current Challenges
Tensor Storage Format
Deep Lake System Overview
Machine Learning Use Cases
Performance Benchmarks
Discussion and Limitations
Related Work
Conclusions, Acknowledgement, and References ABSTRACT Traditional data lakes provide critical data infrastructure for analytical workloads by enabling time travel, running SQL queries, ingesting data with ACID transactions, and visualizing petabytescale datasets on cloud storage. They allow organizations to break down data silos, unlock data-driven decision-making, improve operational efficiency, and reduce costs. However, as deep learning usage increases, traditional data lakes are not well-designed for applications such as natural language processing (NLP), audio processing, computer vision, and applications involving non-tabular datasets. This paper presents Deep Lake, an open-source lakehouse for deep learning applications developed at Activeloop[1][2]. Deep Lake maintains the benefits of a vanilla data lake with one key difference: it stores complex data, such as images, videos, annotations, as well as tabular data, in the form of tensors and rapidly streams the data over the network to (a) Tensor Query Language, (b) in-browser visualization engine, or (c) deep learning frameworks without sacrificing GPU utilization. Datasets stored in Deep Lake can be accessed from PyTorch [58], TensorFlow [25], JAX [31], and integrate with numerous MLOps tools. KEYWORDS - Deep Lake, Deep Learning, Data Lake, Lakehouse, Cloud Computing, Distributed Systems 1. INTRODUCTION A data lake is a central repository that allows organizations to store structured, unstructured, and semi-structured data in one place. Data lakes provide a better way to manage, govern, and analyze data. In addition, they provide a way to break data silos and gain insights previously hidden in disparate data sources. First-generation data lakes traditionally collected data into distributed storage systems such as HDFS [71] or AWS S3 [1]. Unorganized collections of the data turned data lakes into "data swamps", which gave rise to the second-generation data lakes led by Delta, Iceberg, and Hudi [27, 15, 10]. They strictly operate on top of standardized structured formats such as Parquet, ORC, Avro [79, 6, 20] and provide features like time travel, ACID transactions, and schema evolution. Data lakes directly integrate with query engines such as Presto, Athena, Hive, and Photon [70, 12, 76, 66] to run analytical queries. Additionally, they connect to frameworks like Hadoop, Spark, and Airflow [14, 82, 9] for ETL pipeline maintenance. In its turn, the integration between data lakes and query engines with clear compute and storage separation resulted in the emergence of systems like Lakehouse [28] that serve as an alternative to data warehouses, including Snowflake, BigQuery, Redshift, and Clickhouse [33, 4, 40, 2]. Over the past decade, deep learning has outpaced traditional machine learning techniques involving unstructured and complex data such as text, images, videos, and audio [44, 47, 38, 83, 51, 30, 63, 56]. Not only did deep learning systems outgrow traditional techniques, but they also achieved super-human accuracy in applications such as cancer detection from X-ray images, anatomical reconstruction of human neural cells, playing games, driving cars, unfolding proteins, and generating images [61, 48, 72, 42, 77]. Large language models with transformer-based architectures achieved state-of-the-art results across translation, reasoning, summarization, and text completion tasks [78, 36, 81, 32]. Large multi-modal networks embed unstructured data into vectors for cross-modal search [29, 60]. Moreover, they are used to generate photo-realistic images from text [62, 65]. Although one of the primary contributors to the success of deep learning models has been the availability of large datasets such as CoCo (330K images), ImageNet (1.2M images), Oscar (multilingual text corpus), and LAION (400M and 5B images) [49, 34, 74, 68], it does not have a well-established data infrastructure blueprint similar to traditional analytical workloads to support such scale. On the other hand, Modern Data Stack (MDS) lacks the features required to deploy performant deep learning-based solutions so organizations opt to develop in-house systems. In this paper, we introduce Deep Lake, a lakehouse specialized for deep learning workloads. Deep Lake retains the main benefits of a traditional data lake with one notable distinction: it stores complex data, such as images, videos, annotations, and tabular data, as tensors and rapidly streams the data to deep learning frameworks over the network without sacrificing GPU utilization. Furthermore, it provides native interoperability between deep learning frameworks such as PyTorch, TensorFlow, and JAX [58, 25, 31]. The main technical contributions of this paper include: • Tensor Storage Format that stores dynamically shaped arrays on object storage; • Streaming Dataloader that schedules fetching, decompression, and user-defined transformations, optimizing data transfer throughput to GPUs for deep learning; • Tensor Query Language running SQL-like operations on top of multi-dimensional array data; • In-browser visualization engine that streams data from object storage and renders it in the browser using WebGL. The remainder of this paper unfolds as follows. We begin by considering current challenges in deep learning on unstructured data. Next, we present the Tensor Storage Format (TSF) with its key concepts. Furthermore, we discuss Deep Lake’s capabilities and applications within the ML cycle. Next, we provide performance experiments and discuss the results. Finally, we review related work, list possible limitations, and conclude. This paper is available on arxiv under CC 4.0 license. [1] Source code available: https://github.com/activeloopai/deeplake [2] Documentation available at https://docs.deeplake.ai Authors: (1) Sasun Hambardzumyan, Activeloop, Mountain View, CA, USA; (2) Abhinav Tuli, Activeloop, Mountain View, CA, USA; (3) Levon Ghukasyan, Activeloop, Mountain View, CA, USA; (4) Fariz Rahman, Activeloop, Mountain View, CA, USA;. (5) Hrant Topchyan, Activeloop, Mountain View, CA, USA; (6) David Isayan, Activeloop, Mountain View, CA, USA; (7) Mark McQuade, Activeloop, Mountain View, CA, USA; (8) Mikayel Harutyunyan, Activeloop, Mountain View, CA, USA; (9) Tatevik Hakobyan, Activeloop, Mountain View, CA, USA; (10) Ivo Stranic, Activeloop, Mountain View, CA, USA; (11) Davit Buniatyan, Activeloop, Mountain View, CA, USA. Authors: Authors: (1) Sasun Hambardzumyan, Activeloop, Mountain View, CA, USA; (2) Abhinav Tuli, Activeloop, Mountain View, CA, USA; (3) Levon Ghukasyan, Activeloop, Mountain View, CA, USA; (4) Fariz Rahman, Activeloop, Mountain View, CA, USA;. (5) Hrant Topchyan, Activeloop, Mountain View, CA, USA; (6) David Isayan, Activeloop, Mountain View, CA, USA; (7) Mark McQuade, Activeloop, Mountain View, CA, USA; (8) Mikayel Harutyunyan, Activeloop, Mountain View, CA, USA; (9) Tatevik Hakobyan, Activeloop, Mountain View, CA, USA; (10) Ivo Stranic, Activeloop, Mountain View, CA, USA; (11) Davit Buniatyan, Activeloop, Mountain View, CA, USA. Table of Links Abstract and Intro Current Challenges Tensor Storage Format Deep Lake System Overview Machine Learning Use Cases Performance Benchmarks Discussion and Limitations Related Work Conclusions, Acknowledgement, and References Abstract and Intro Abstract and Intro Current Challenges Current Challenges Tensor Storage Format Tensor Storage Format Deep Lake System Overview Deep Lake System Overview Machine Learning Use Cases Machine Learning Use Cases Performance Benchmarks Performance Benchmarks Discussion and Limitations Discussion and Limitations Related Work Related Work Conclusions, Acknowledgement, and References Conclusions, Acknowledgement, and References ABSTRACT Traditional data lakes provide critical data infrastructure for analytical workloads by enabling time travel, running SQL queries, ingesting data with ACID transactions, and visualizing petabytescale datasets on cloud storage. They allow organizations to break down data silos, unlock data-driven decision-making, improve operational efficiency, and reduce costs. However, as deep learning usage increases, traditional data lakes are not well-designed for applications such as natural language processing (NLP), audio processing, computer vision, and applications involving non-tabular datasets. This paper presents Deep Lake, an open-source lakehouse for deep learning applications developed at Activeloop[1][2]. Deep Lake maintains the benefits of a vanilla data lake with one key difference: it stores complex data, such as images, videos, annotations, as well as tabular data, in the form of tensors and rapidly streams the data over the network to (a) Tensor Query Language, (b) in-browser visualization engine, or (c) deep learning frameworks without sacrificing GPU utilization. Datasets stored in Deep Lake can be accessed from PyTorch [58], TensorFlow [25], JAX [31], and integrate with numerous MLOps tools. KEYWORDS - Deep Lake, Deep Learning, Data Lake, Lakehouse, Cloud Computing, Distributed Systems 1. INTRODUCTION A data lake is a central repository that allows organizations to store structured, unstructured, and semi-structured data in one place. Data lakes provide a better way to manage, govern, and analyze data. In addition, they provide a way to break data silos and gain insights previously hidden in disparate data sources. First-generation data lakes traditionally collected data into distributed storage systems such as HDFS [71] or AWS S3 [1]. Unorganized collections of the data turned data lakes into "data swamps", which gave rise to the second-generation data lakes led by Delta, Iceberg, and Hudi [27, 15, 10]. They strictly operate on top of standardized structured formats such as Parquet, ORC, Avro [79, 6, 20] and provide features like time travel, ACID transactions, and schema evolution. Data lakes directly integrate with query engines such as Presto, Athena, Hive, and Photon [70, 12, 76, 66] to run analytical queries. Additionally, they connect to frameworks like Hadoop, Spark, and Airflow [14, 82, 9] for ETL pipeline maintenance. In its turn, the integration between data lakes and query engines with clear compute and storage separation resulted in the emergence of systems like Lakehouse [28] that serve as an alternative to data warehouses, including Snowflake, BigQuery, Redshift, and Clickhouse [33, 4, 40, 2]. Over the past decade, deep learning has outpaced traditional machine learning techniques involving unstructured and complex data such as text, images, videos, and audio [44, 47, 38, 83, 51, 30, 63, 56]. Not only did deep learning systems outgrow traditional techniques, but they also achieved super-human accuracy in applications such as cancer detection from X-ray images, anatomical reconstruction of human neural cells, playing games, driving cars, unfolding proteins, and generating images [61, 48, 72, 42, 77]. Large language models with transformer-based architectures achieved state-of-the-art results across translation, reasoning, summarization, and text completion tasks [78, 36, 81, 32]. Large multi-modal networks embed unstructured data into vectors for cross-modal search [29, 60]. Moreover, they are used to generate photo-realistic images from text [62, 65]. Although one of the primary contributors to the success of deep learning models has been the availability of large datasets such as CoCo (330K images), ImageNet (1.2M images), Oscar (multilingual text corpus), and LAION (400M and 5B images) [49, 34, 74, 68], it does not have a well-established data infrastructure blueprint similar to traditional analytical workloads to support such scale. On the other hand, Modern Data Stack (MDS) lacks the features required to deploy performant deep learning-based solutions so organizations opt to develop in-house systems. In this paper, we introduce Deep Lake, a lakehouse specialized for deep learning workloads. Deep Lake retains the main benefits of a traditional data lake with one notable distinction: it stores complex data, such as images, videos, annotations, and tabular data, as tensors and rapidly streams the data to deep learning frameworks over the network without sacrificing GPU utilization. Furthermore, it provides native interoperability between deep learning frameworks such as PyTorch, TensorFlow, and JAX [58, 25, 31]. The main technical contributions of this paper include: • Tensor Storage Format that stores dynamically shaped arrays on object storage; • Tensor Storage Format • Streaming Dataloader that schedules fetching, decompression, and user-defined transformations, optimizing data transfer throughput to GPUs for deep learning; • Streaming Dataloader • Tensor Query Language running SQL-like operations on top of multi-dimensional array data; • Tensor Query Language • In-browser visualization engine that streams data from object storage and renders it in the browser using WebGL. • In-browser visualization engine The remainder of this paper unfolds as follows. We begin by considering current challenges in deep learning on unstructured data. Next, we present the Tensor Storage Format (TSF) with its key concepts. Furthermore, we discuss Deep Lake’s capabilities and applications within the ML cycle. Next, we provide performance experiments and discuss the results. Finally, we review related work, list possible limitations, and conclude. This paper is available on arxiv under CC 4.0 license. This paper is available on arxiv under CC 4.0 license. available on arxiv [1] Source code available: https://github.com/activeloopai/deeplake [2] Documentation available at https://docs.deeplake.ai

Deep Lake, a Lakehouse for Deep Learning: Abstract and Intro

ডিপ লেক, গভীর শিক্ষার জন্য একটি লেকহাউস: বিমূর্ত এবং ভূমিকা

Deep Lake, ein Lakehouse für Deep Learning: Zusammenfassung und Einführung

Deep Lake, una casa de lago para el aprendizaje profundo: resumen e introducción

Deep Lake, un Lakehouse pour le Deep Learning : résumé et introduction

डीप लेक, डीप लर्निंग के लिए एक लेकहाउस: सार और परिचय

ディープラーニングのためのレイクハウス、Deep Lake: 概要と紹介

딥 러닝을 위한 레이크하우스, Deep Lake: 개요 및 소개

Deep Lake, uma casa de lago para aprendizado profundo: resumo e introdução

Deep Lake, домик у озера для глубокого обучения: Аннотация и введение

Deep Lake, Derin Öğrenme için Bir Göl Evi: Özet ve Giriş

Deep Lake, Lakehouse cho Deep Learning: Tóm tắt và giới thiệu

Deep Lake，深度学习的 Lakehouse：摘要和简介

Stories, News & Mentions about CANDER

Cander’s Stories on HackerNoon