WeArisma

wearisma

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

Odd thought - climate change is real but accelerated by the appearance of humans and industrialization. We see the effects and the extremities because we're a global species with the technologies to view and record it happening. No other species did in the past, so they were never aware of it, nor the great extinction events that were taking place. These took millions of years, and it's still incomprehensible for humans to think along those timeframes. Change on a planetary scale isn't going to be altered by some man-made doohickery or planting trees; it's an unstoppable force of nature at work that will last beyond human civilization. The chances are that every species that has lived or will ever live on Earth has and will be a part of this cycle, and we're just one of them with the capacity to grasp what is happening and the audacity to believe we can stop it. In a million years, the surface of the Earth could be an arid wasteland with oceans choked with plastic and an atmosphere that isn't breathable. It could take another hundred million years to become habitable again with new life that might not be human anymore. Who knows? No-one does. The point is, what we think we can do now with all our fancy promises about carbon emissions in 2050 and technologies we can chat to will have absolutely no effect on a chain of events that takes thousands upon thousands of years to unfold. We're just not prepared to accept it. Think about the evolutionary cycles the planet has already witnessed in 4 billion years and will witness again. Our involvement has only hurried some of that along by wiping out species by our own actions, but arguably, they could have been wiped out anyway. Or maybe that species would have been the seed for another evolutionary cycle that would have driven life to emerge again, and in destroying them, we've removed the chance for life to survive this next cataclysm. The Earth ends up like Mars. What’s also pretty clear is that the “Climate Scam” people are partially correct, but not in the way they believe. The Truth About Climate Change There is climate change. It's coming. It's happening and will continue to happen over millennia. The real 'scam' is the narrative that any human impact can be mitigated with political will and increased taxation to slow/ reverse/ halt/ alter the trajectory of something this big. It's sheer nonsense. As is the Stop Oil bullshit performance art. Not even science can explain what happened in the past accurately as much as it can project accurately what will happen over the next 10 million years. Plastics can take anywhere from 20 to 500 years to decompose, depending on the material and structure, and it takes 1,000 years for a plastic bag to degrade in a landfill. Unfortunately, the bags don't break down completely but instead photo-degrade, becoming microplastics that absorb toxins and continue to pollute the environment. But even in that timeframe, that's just yet another blip in terms of how long this planet's climate will take to recycle to its eventual conclusion and bring around another extinction and evolutionary beginning as it has done many times before. The climate tech market that is worth $20bn is absolutely meaningless, nothing more than feel-good technologies to ease the troubled mind and make it appear we’re doing something impactful. There is no impact here, no moving of a planetary-sized needle by tiny hands. We will try; we can’t even claim that we’ll try our best because we don’t. We do enough to signal that we made an effort and no more, all the while burning more resources to make products and technologies that really do fuck all in the grand scheme of things. Venture capitalists who claim to look for exponential ideas or concepts to accelerate humanity really don’t stand by any of it. Gotta love VCs whose investment thesis revolves around funding technologies that accelerate humanity while civilization is on the brink thanks to the collapse of the climate caused by funding technologies that accelerate humanity. The chances are, that when this extinction event is finished, there will be nothing left to show we existed, no wonderous technologies or buildings, nor anything in terms of new life that will emerge that can make sense of any of it. It’s funny to think that humanoids are at the top of the evolutionary process and possess the ability to harness the world around them, even funnier to consider that they all get one shot with the planet they inhabit. The Fermi Paradox may well be explained not just by distance and time but by the fact that every life-bearing planet went through several evolutionary extinctions to produce a humanoid species capable of understanding and intelligence only for them to fuck it all up every single time. Who knows, there may well be an infinitely small percentage that makes it, that don’t make the same mistakes we do and are, like us at this moment in time, searching for themselves elsewhere in the Universe. Read World Without Us by Alan Weisman then go live your best life. The dinosaurs certainly did. Also published here

Climate Change: There Is Nothing You Can Do About It - Part 1

Emphatic words must stand in emphatic positions

HOLLINGFORD GOSSIPS.

Now—soft awhile—have I arrived so near the end?

On the Economy of Machinery and Manufactures by Charles Babbage is part of the HackerNoon Books Series. You can jump to any chapter in this book . Registering Operations here Chapter 8. Registering Operations One great advantage which we may derive from machinery is from the check which it affords against the inattention, the idleness, or the dishonesty of human agents. Few occupations are more wearisome than counting a series of repetitions of the same fact; the number of paces we walk affords a tolerably good measure of distance passed over, but the value of this is much enhanced by possessing an instrument, the pedometer, which will count for us the number of steps we have made. A piece of mechanism of this kind is sometimes applied to count the number of turns made by the wheel of a carriage, and thus to indicate the distance travelled: an instrument, similar in its object, but differing in its construction, has been used for counting the number of strokes made by a steam-engine, and the number of coins struck in a press. One of the simplest instruments for counting any series of operations, was contrived by Mr Donkin.(1*) Another instrument for registering is used in some establishments for calendering and embossing. Many hundred thousand yards of calicoes and stuffs undergo these operations weekly; and as the price paid for the process is small, the value of the time spent in measuring them would bear a considerable proportion to the profit. A machine has, therefore, been contrived for measuring and registering the length of the goods as they pass rapidly through the hands of the operator, by which all chance of erroneous counting is avoided. Perhaps the most useful contrivance of this kind, is one for ascertaining the vigilance of a watchman. It is a piece of mechanism connected with a clock placed in an apartment to which the watchman has not access; but he is ordered to pull a string situated in a certain part of his round once in every hour. The instrument, aptly called a tell-tale, informs the owner whether the man has missed any, and what hours during the night. It is often of great importance, both for regulations of excise as well as for the interest of the proprietor, to know the quantity of spirits or of other liquors which have been drawn off by those persons who are allowed to have access to the vessels during the absence of the inspectors or principals. This may be accomplished by a peculiar kind of stop-cock—which will, at each opening, discharge only a certain measure of fluid the number of times the cock has been turned being registered by a counting apparatus accessible only to the master. The time and labour consumed in gauging the contents of casks partly filled, has led to an improvement which, by the simplest means, obviates a considerable inconvenience, and enables any person to read off, on a scale, the number of gallons contained in any vessel, as readily as he does the degree of heat indicated by his thermometer. A small stop-cock connects the bottom of the cask with a glass tube of narrow bore fixed to a scale on the side of the cask, and rising a little above its top. The plug of the cock may be turned into three positions: in the first, it cuts off all communication with the cask: in the second, it opens a communication between the cask and the glass tube: and, in the third. It cuts off the connection between the cask and the tube, and opens a communication between the tube and any vessel held beneath the cock to receive its contents. The scale of the tube is graduated by pouring into the cask successive quantities of water, while the communication between the cask and the tube is open. Lines are then drawn on the scale opposite the places in the tube to which the water rises at each addition, and the scale being thus formed by actual measurement,(2*) the contents of each cask are known by inspection, and the tedious process of gauging is altogether dispensed with. Other advantages accrue from this simple contrivance, in the great economy of time which it introduces in making mixtures of different spirits, in taking stock, and in receiving spirit from the distiller. The gas-meter, by which the quantity of gas used by each consumer is ascertained, is another instrument of this kind. They are of various forms, but all of them intended to register the number of cubic feet of gas which has been delivered. It is very desirable that these meters should be obtainable at a moderate price, and that every consumer should employ them; because, by making each purchaser pay only for what he consumes, and by preventing that extravagant waste of gas which we frequently observe, the manufacturer of gas will be enabled to make an equal profit at a diminished price to the consumer. The sale of water by the different companies in London, might also, with advantage, be regulated by a meter. If such a system were adopted, much water which is now allowed to run to waste would be saved, and an unjust inequality between the rates charged on different houses by the same company be avoided. Another most important object to which a meter might be applied, would be to register the quantity of water passing into the boilers of steam-engines. Without this, our knowledge of the quantity evaporated by different boilers, and with fireplaces of different constructions, as well as our estimation of the duty of steam-engines, must evidently be imperfect. Another purpose to which machinery for registering operations is applied with much advantage is the determination of the average effect of natural or artificial agents. The mean height of the barometer, for example, is ascertained by noting its height at a certain number of intervals during the twenty-four hours. The more these intervals are contracted, the more correctly will the mean be ascertained; but the true mean ought to be influenced by each momentary change which has occurred. Clocks have been proposed and made with this object, by which a sheet of paper is moved, slowly and uniformly, before a pencil fixed to a float upon the surface of the mercury in the cup of the barometer. Sir David Brewster proposed, several years ago to suspend a barometer, and swing it as a pendulum. The variations in the atmosphere would thus alter the centre of oscillation, and the comparison of such an instrument with a good clock, would enable us to ascertain the mean altitude of the barometer during any interval of the observer's absence.(3*) An instrument for measuring and registering the quantity of rain, was invented by Mr John Taylor, and described by him in the Philosophical Magazine. It consists of an apparatus in which a vessel that receives the rain falling into the reservoir tilts over as soon as it is full, and then presents another similar vessel to be filled, which in like manner, when full, tilts the former one back again. The number of times these vessels are emptied is registered by a train of wheels; and thus, without the presence of the observer, the quantity of rain falling during a whole year may be measured and recorded. Instruments might also be contrived to determine the average force of traction of horses—of the wind—of a stream or of any irregular and fluctuating effort of animal or other natural force. Clocks and watches may be considered as instruments for registering the number of vibrations performed by a pendulum or a balance. The mechanism by which these numbers are counted is technically called a scapement. It is not easy to describe: but the various contrivances which have been adopted for this purpose, are amongst the most interesting and most ingenious to which mechanical science has given birth. Working models, on an enlarged scale, are almost necessary to make their action understood by the unlearned reader; and, unfortunately, these are not often to be met with. A very fine collection of such models exists amongst the collection of instruments at the University of Prague. Instruments of this kind have been made to extend their action over considerable periods of time, and to register not merely the hour of the day, but the days of the week, of the month, of the year, and also to indicate the occurrence of several astronomical phenomena. Repeating clocks and watches may be considered as instruments for registering time, which communicate their information only when the owner requires it, by pulling a string, or by some similar application. An apparatus has recently been applied to watches, by which the hand which indicates seconds leaves a small dot of ink on the dial-plate whenever a certain stop or detent is pushed in. Thus, whilst the eye is attentively fixed on the phenomenon to be observed, the finger registers on the face of the watch-dial the commencement and the end of its appearance. Several instruments have been contrived for awakening the attention of the observer at times previously fixed upon. The various kinds of alarums connected with clocks and watches are of this kind. In some instances it is desirable to be able to set them so as to give notice at many successive and distant points of time, such as those of the arrival of given stars on the meridian. A clock of this kind is used at the Royal Observatory at Greenwich. An earthquake is a phenomenon of such frequent occurrence, and so interesting, both from its fearful devastations as well as from its connection with geological theories, that it becomes important to possess an instrument which shall, if possible, indicate the direction of the shock, as well as its intensity. An observation made a few years since at Odessa, after an earthquake which happened during the night, suggests a simple instrument by which the direction of the shock may be determined. A glass vase, partly filled with water, stood on the table of a room in a house at Odessa; and, from the coldness of the glass, the inner part of the vessel above the water was coated with dew. Several very perceptible shocks of an earthquake happened between three and four o'clock in the morning; and when the observer got up, he remarked that the dew was brushed off at two opposite sides of the glass by a wave which the earthquake had caused in the water. The line joining the two highest points of this wave was, of course, that in which the shock travelled. This circumstance, which was accidentally noticed by an engineer at Odessa,(4*) suggests the plan of keeping, in countries subject to earthquakes, glass vessels partly filled with treacle, or some unctuous fluid, so that when any lateral motion is communicated to them from the earth, the adhesion of the liquid to the glass shall enable the observer, after some interval of time, to determine the direction of the shock. In order to obtain some measure of the vertical oscillation of the earth, a weight might be attached to a spiral spring, or a pendulum might be sustained in a horizontal position, and a sliding index be moved by either of them, so that the extreme deviations should be indicated by it. This, however, would not give even the comparative measure accurately, because a difference in the velocity of the rising or falling of the earth's surface would affect the instrument. NOTES: Transactions of the Society of Arts, 1819, p. 116. The contrivance is due to Mr Hencky, of High Holborn, in whose establishment it is in constant use. About seven or eight years since, without being aware of Sir David Brewster's proposal. I adapted a barometer, as a pendulum, to the works of a common eight day clock: it remained in my library for several months, but I have mislaid the observations which were made. Memoires de l'Academie des Sciences de Petersburgh, 6e serie, tom. i. p. 4. About HackerNoon Book Series: We bring you the most important technical, scientific, and insightful public domain books. This book is part of the public domain. Charles Babbage (2003). On the Economy of Machinery and Manufactures. Urbana, Illinois: Project Gutenberg. Retrieved https://www.gutenberg.org/cache/epub/4238/pg4238.html This eBook is for the use of anyone anywhere at no cost and with almost no restrictions whatsoever. You may copy it, give it away or re-use it under the terms of the Project Gutenberg License included with this eBook or online at , located at . www.gutenberg.org https://www.gutenberg.org/policy/license.html

Registering Operations

THE MORTAL IMMORTAL

MAKING FRIENDSHIP.

How it Works by Archibald Williams is part of the HackerNoon Books Series. You can jump to any chapter in this book . PREFACE here PREFACE. How does it work? This question has been put to me so often by persons young and old that I have at last decided to answer it in such a manner that a much larger public than that with which I have personal acquaintance may be able to satisfy themselves as to the principles underlying many of the mechanisms met with in everyday life. In order to include steam, electricity, optics, hydraulics, thermics, light, and a variety of detached mechanisms which cannot be classified under any one of these heads, within the compass of about 450 pages, I have to be content with a comparatively brief treatment of each subject. This brevity has in turn compelled me to deal with principles rather than with detailed descriptions of individual devices—though in several cases recognized types are examined. The reader will look in vain for accounts of the Yerkes telescope, of the latest thing in motor cars, and of the largest locomotive. But he will be put in the way of understanding the essential nature of telescopes, motors, and steam-engines so far as they are at present developed, which I think may be of greater ultimate profit to the uninitiated. all While careful to avoid puzzling the reader by the use of mysterious phraseology I consider that the parts of a machine should be given their technical names wherever possible. To prevent misconception, many of the diagrams accompanying the letterpress have words as well as letters written on them. This course also obviates the wearisome reference from text to diagram necessitated by the use of solitary letters or figures. I may add, with regard to the diagrams of this book, that they are purposely somewhat unconventional, not being drawn to scale nor conforming to the canons of professional draughtsmanship. Where advisable, a part of a machine has been exaggerated to show its details. As a rule solid black has been preferred to fine shading in sectional drawings, and all unnecessary lines are omitted. I would here acknowledge my indebtedness to my draughtsman, Mr. Frank Hodgson, for his care and industry in preparing the two hundred or more diagrams for which he was responsible. Four organs of the body—the eye, the ear, the larynx, and the heart—are noticed in appropriate places. The eye is compared with the camera, the larynx with a reed pipe, the heart with a pump, while the ear fitly opens the chapter on acoustics. The reader who is unacquainted with physiology will thus be enabled to appreciate the better these marvellous devices, far more marvellous, by reason of their absolutely automatic action, than any creation of human hands. A.W. Uplands, Stoke Poges, Bucks. About HackerNoon Book Series: We bring you the most important technical, scientific, and insightful public domain books. This book is part of the public domain. Archibald Williams (2009). How it Works. Urbana, Illinois: Project Gutenberg. Retrieved https://www.gutenberg.org/cache/epub/28553/pg28553-images.html This eBook is for the use of anyone anywhere at no cost and with almost no restrictions whatsoever. You may copy it, give it away or re-use it under the terms of the Project Gutenberg License included with this eBook or online at , located at . www.gutenberg.org https://www.gutenberg.org/policy/license.html

How does it work?

Stories, News & Mentions about WEARISMA

Wearisma’s Stories on HackerNoon