Just learned AI can now detect Parkinson's from a breathing pattern study I saw on the NIH site

I was reading through the NIH database last week looking for something completely different and stumbled on this MIT study about how they trained an AI model to spot Parkinson's disease just from analyzing someone's breathing while they sleep. It hit me because my uncle has Parkinson's and they caught it late after years of symptoms getting brushed off as just aging stuff. The model apparently worked with 86% accuracy in a group of nearly 800 people which is wild compared to how doctors usually spot it. I guess they used a neural network trained on breathing patterns from a device you just wear on your chest at night. Has anyone else come across this kind of early detection AI stuff for other conditions?

2 comments

2 Comments

the_wesley22d ago

Whoa, I gotta respectfully disagree with the hype here @the_kim. I mean sure, 86% accuracy sounds impressive, but that still means one in every seven people gets a false result (and that's just the ones they published). Plus, a device you wear on your chest at night? That's a huge barrier for most folks who aren't already in a sleep study. Early detection is cool in theory, but I'm not convinced this beats a simple clinical exam by a neurologist who knows what to look for.

the_kim23d ago

Crazy timing seeing this, I literally just sent a similar MIT article to my dad because his sleep patterns changed years before his Parkinson's diagnosis and nobody connected the dots. Are there any other conditions they're hitting with this kind of breathing analysis?