Recognizing the Opportunities for Advanced Computing in Healthcare
The two topics of healthcare and technology have truly joined forces together to create a revolutionary era for healthcare today. I studied Biochemistry as one of my double majors at UC Berkeley many moons ago but decided to enter the field of Computer Science & application development when I entered the marketplace. At the time breakthroughs surrounding healthcare weren’t happening fast enough but now, with the combination of advancements in genetics, medical research, and the democratization of advanced data computing techniques, we are getting a glimpse into the future possibilities in improving healthcare substantially.
Why it Hasn’t Happened Yet
There is a multitude of reasons that have contributed to why AI hasn’t massively transformed healthcare yet. I think these are some of the biggest hurdles:
- Slow Advanced Technology Adoption in Healthcare
Considering the marturity of AI & ML today, we’ve already made a lot of progress in the field by applying it toward image recognition and automating some health diagnostics. But when it comes to leveraging AI for more advanced treatment recommendations, there is still ground to cover. As of now, we are still waiting for the maturation of the newer and better deep learning frameworks and its models, the simplification of the user experience surrounding its use, and not to mention the lack of comprehensive understanding on cause and factors that contribute to certain illnesses, in many instances due to lack of robust data.
2. Simplify Data Collection Process to Capture Detail Granularity:
Another factor is the quality and extraction process of the available data sources. We have a multitude of different data sources, but data such as electronic medical records are still very coarse and slow. For example, manual transcribing or typing out each medical procedure is still being practiced even though we have technology that can very accurately automate the speech-to-text and video-to-text conversion and create auto-relationships between different data elements through a graph. We can improve patient care by making more data and more granular data capture easy — and the technology to do so exists and is more than capable enough to do so today.
3. Complexities & Mysteries of Human Health:
The third intricacy is us — humans. We are anatomically complex living beings, so while two individuals who may have the same diagnosis, they may still react differently to the same treatment plan based on their different surrounding environment variables. With so many variables present, our AI needs to have access to widest and diverse datasets possible that doesn’t force a litmited finding.
Promising Use Cases
With the current reliability of AI technology, there are many low hanging fruits to pick for practical implementations that can actually impact healthcare in a very fundamental way, for example,
- Increase the Accessibility of AI-based Diagnostics to the Masses:
We’ve had a lot of success applying image recognition in high tech manufacturing, discrete manufacturing to help identify product defects. I don’t see a viable reason why the same technology cannot be applied to radiology information or MRI data where we’re looking at variants. More over, we can take this same technology to the masses. We have just starting to see the possibility of a user sending a picture via mobile phone, connecting to a cloud API of a pre-trained AI model to give a quick diagnosis. There are many easy methods opening up as opportunities to leverage image recognition technology where one can determine and classify from a variety of medical diagnoses ranging from analyzing ear patterns for problems with the ear, nose or throat to whether a skin lesion is a benign skin lesion or a malignant skin cancer to if a skin rash is an acne, chicken pox or sunburn. All that one requires to get a augmented and automated diagnosis is taking a picture. Think of the applicability of this to all the parents with little ones (myself included).
2. Data-rich, Electronic Health Record (EHR):
The mean rate of manual documentation in most hospitals can range between 12% to 35% in a physician or nurses’ regular shift. Today, we have the technology to auto-transcribe voice-recorded conversations between doctors and patients directly into their health record, or even transcribing what’s been captured in a video recording. Instead of typing a report later, we can transcribe a video or voice recording of a procedure that a doctor has performed right there and then. Technologies as such could additionally make medical research easier just by making the data more accessible.
3. Treatment Recommendations:
AI is nearly ready for healthcare when it comes to creating extremely precise analytical technology-based images not solely just for diagnosis but to make advanced analytical treatment recommendations as well. This concept is, however, more for a future opportunity — I don’t believe it is prime time yet as this is a very difficult problem to address with AI mainly due to the issues that lie largely outside of the technology domain.
The Role of Wearable Technology and the Rise of Opted-in Open Data Sharing Network
The role of tracking apps or wearable technology isn’t just for the sole purpose of datasets, it can also be an advancement for individuals in taking control of their own health. For example, some wearable technologies have an EKG (electrocardiography) capability like the latest Apple Watch. Why is that important? People today are taught to look for early warnings of heart attack symptoms, but we don’t always accurately translate knowledge into action. I, unfortunately, had a friend who is no longer with us because he didn’t seek emergency treatment when the first sign presented itself. On the other hand, 8 million patients a year in the United States are being falsely administered to hospitals for fear they will have a heart attack. A device that can look at baselines versus variances can help with the recognition of what is happening during that particular situation.
As we commoditize these wearable technologies, we also need to support people in making their own datasets available. In addition to the physical states tracked by wearable technology, many people, like me, use an app to track our workouts, diets, including micro and macronutrients. These are valuable datasets, that could be great for the advancement of medical and population health research and diagnosis if people make it available. Encouraging this kind of behavior would really help to further advance the medical sciences and healthcare research leading us to see the next breakthrough in healthcare improvement.
I truly believe the popularity of wearable technologies like the Apple watch or Garmins, will be the entry point to accelerate AI’s adoption in healthcare. I personally love to wear these, and the collection of all such valuable information from individuals gives us access to data that we previously couldn’t even imagine having. The convergence of these research areas and the possibilities of increased personal data will bring us to the nexus of transformation within the healthcare industry.
Platforms for the Healthcare Revolution
The best technologies in data platform and data computing are ready to make all of the possibilities discussed above happen. Today, the capability exists to analyze multiple data types, whether its image, voice recording, text document, or computer records. There are so many computing techniques available beyond AI that we have access to like advanced analytics capability, geospatial, text analysis, and natural language processing. Consider the possibilities of being able to analyze medical records across 27 different languages all of which we can do today.
Healthcare as an industry has some of the most diverse data types coming together — a mass amount of data just waiting to explode. I’m confident that the modern, integrated data platform technology is ready to meet the challenges of a healthcare revolution.