How will artificial intelligence impact healthcare?
Artificial intelligence has been touted as the cure to our healthcare crisis for a couple decades now. Recently there has been some progress made with diagnosing certain conditions (diabetic retinopathy/eye changes in diabetes, otitis media/ear infection and cancer detection). Even with the recent advancements, adoption remains spotty and any actionable steps continue to require considerable physician input.
In a previous podcast with Amit Garg from Tau Ventures, we discussed the concept of explainability in the realm of artificial intelligence. For a long time, we held on to the notion that we need to understand the learning processes (neural networks, decision trees, linear regression) behind artificial intelligence. While I am far from an expert on this, a neural network (one type of machine learning) is essentially a black box. If we require explainability as a precursor to adoption (FDA approval in healthcare) then we limit the scope of AI immensely.
I recently spoke with Amit Mehta from BuildersVC, he has considerable experience with artificial intelligence as it pertains to the FDA approval process. The FDA has given approval to several imaging technologies which help physicians in decision making. Here are some takeaways from our discussion.
Identifying potential diagnosis is version 1 for AI. This version answers the question, eg. "is this cancer?"
Version 2 will lead to predicting prognosis, if version 1 says “Is it cancer or not right now”, version 2 will say “Will it ever become cancer”. Currently we are at version 2.
Version 3 involves A.I making clinical decisions. Using the above example, version 3 might say “This mass is cancer, it requires a biopsy within 2 weeks.”
Zooming out and looking at our system as a whole, there is an immediate need and opportunity for version 3 AI replacing our algorithmic decisions. An example could be a prescription refill for birth control. This is a very exciting time to be in the healthcare space, we predict the next 5 years will bring on version 3 AI, thereby improving access to care on a global scale.
Thanks for reading,
Rishad
