Grateful for the opportunity to testify today at @congressdotgov on a path forward for safe adoption of AI in healthcare. The potential upside is enormous, but we need the right safeguards and processes to capture the upside efficiently without unforced errors.

A decade ago Viz.ai started as an imaging detection tool. Over 10 years, it has grown to so much more, and it feels like we're adding new capabilities all the time. More imaging algos, automated measures, explainability maps, transfer activation (including helicopters!), chart-aware clinical decision support grounded in society guidelines... and my favorite? Condition-specific chart summarization. Imagine walking into your clinic with a chief-level resident summary for every patient? And their care gaps from society guidelines already flagged for you? Once a dream, now standard across the platform. Dr Chris Mansi and I wrote about it in NEJM AI a year ago as a vision lnkd.in/gR_vnD3z , and now it's reality.

Where can healthcare AI actually be deflationary? Without jumping all the way to autonomous AI dropping unit costs (which is possible, but also not inevitable), I show examples here where AI can be both be incentivized and reduce episode spending. In fact, we have an AI example almost a decade old...

“Unfortunately, this story is not unique to sub-types of heart failure, or even cardiology more broadly. You can find it in every corner of the health care system: a condition presents with symptoms easily mistaken for something far more common, the clinician reasonably treats what it looks like, the patient does not quite get better, and the search quietly stops at a diagnosis that was good enough to move on from. Every clinician — the pulmonologist, the neurologist, the generalist — face their own version of the same challenge. It reflects the modern state of medicine — we have far more diagnoses to choose from, and more inputs to integrate to choose the right one. And it is precisely where AI tools can deliver. But getting them adopted goes far beyond a successful algorithm.” - @NEJM_AI lnkd.in/p/gjwWXiMZ

While there have been calls to slow AI down, I see it differently. Much of it actually needs to go faster. lnkd.in/p/gF6sXs8y

👀