Preprint Announcement – Guide to Regulating Medical AI

One year ago, I left the start-up where I had been working on an AI-driven companion to accompany patients through their cancer treatments.

When I left, I was deeply frustrated with the start-up environment surrounding AI in Healthcare. I was still convinced that AI could help in this space, but all I was seeing was teams going down what I considered to be the wrong paths.

They were vastly over-promising to both their funders and their clients. They were blindly underestimating the difficulties involved. And, they all showed a deep-seated unwillingness to take these problems seriously.

My answer, was to write about the problem. In partnership with Vince Madai, I published an in-depth guide to developing AI in Healthcare Products, first in an Preprint, and later in an open-access peer-reviewed academic journal.

So far, the uptake of our article has been considerable and word is spreading.

However, in writing the article, I discovered that a true regulatory approach to developing medical AI is still lacking. The guidelines are still being written.

In order to contribute to this process, I have written my own solution. It is now available, in an early-draft format, again from a Preprint server.

The idea is simple. These products need to be regulated, but very few people in the world know what a good regulatory package might look like. My article goes through the building blocks for such a package. You will be surprised by just how many steps you need to complete before you can be considered to have a safe product. There are no short-cuts in life.

Often when developing medical AI products the data most required is least available. Take the example of patients with a serious medical condition such as potential for acute kidney injury. It is easy to gather data from healthy patients. It is also possible to find plenty of patients who do not survive. But how do you identify and sample from patients who are highly at-risk but still saveable?

Version 1 is a draft. The intention is to have it out now, for the ongoing international guideline development efforts. There will be mistakes. There are a lot of papers which I have not sufficiently cited. I welcome all comments which try to bring this thing forwards. We need clear regulatory standards, so that our engineers can get back to building products!

2 Replies to “Preprint Announcement – Guide to Regulating Medical AI”

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.