AI predicts if and when someone will experience cardiac arrest
hub.jhu.edu
hub.jhu.edu
> Current clinical cardiac image analysis extracts only simple scar features like volume and mass, severely underutilizing what's demonstrated in this work to be critical data.
> "The images carry critical information that doctors haven't been able to access," said first author Dan Popescu, a former Johns Hopkins doctoral student.
This is, like, the thing that machine learning is really good at--pulling out additional information from complex patterns represented as imagery. It's a very similar reasoning to why neural networks can be really great at classifying malware based on the program structure or doing semantic segmentation on well-defined scenes.
> The team trained a second neural network to learn from 10 years of standard clinical patient data, 22 factors such as patients' age, weight, race, and prescription drug use.
Again, this kind of multi-dimensional pattern-matching is pretty much much a machine learning wheelhouse.
> [They] were validated in tests with an independent patient cohort from 60 health centers across the United States, with different cardiac histories and different imaging data
And apparently has a bunch of independent validation. Neat stuff, I'm curious to see how it will develop/be distributed in the future.
Disclosure: I recently accepted a job offer from JHU, albeit in a different field.
But if you were going to apply data science/machine learning to healthcare (personally, I dislike the term "AI"), I think that this project would be very close to both how and where you would want it. Peer-reviewed research with independent confirmation, in an narrow, image-based diagnosis, overseen by medical professionals throughout the process (not directly making recommendations to patients), by one of the world's leading medical research institutions.
This isn't a catch-all, and it's certainly possible that cynicism could be warranted down the line. But for me, personally, I think that cautious optimism for this specific situation overrides the cynicism.
Everything I’ve seen has been about flagging stuff for further human review, what things have you seen that have made you so skeptical?
the big problem with ML in medicine is that the experts of the two fields lack the - in both cases very deep - domain expertise of the other field.
Ask yourself much much top talent in ML there really is, now ask yourself how many of those work on a specific problem or toolset in medicine. Then ask yourself how much top talent in the medical fields has a profound intersection with statistics. Both are deep fields dealing with inherently complicated problem spaces. The amount of people who are able to make profound contributions to both fields can probably fit in a medium-sized conference room, while there are hundreds, if not a magnitude more, ML-driven medical research projects. It's hard enough making contributions to either field without heavily intersecting with a completely different domain, and a lot of the best-in-class experts of ML (or, better: AI (!)) are busy solving things like protein folding in London. And if you look at those guys faces, they look tired, because what they are doing is really, really hard.
This could help women that say so, as well as minorities, and even more so for minority women
No longer would people’s complaints be dismissed as heuristics (edit: hysterics) just because a doctor doesn't understand how people express themselves
at least for predictors of cardiac arrest
Our two daughters both have unknown heart variants and one of them is on a beta blocker for life since the age of 1 to avoid sudden cardiac episodes. She’s also forbidden to drink stimulants or play certain sports for risk of heart episodes. She looks completely healthy and normal and wants to play as intense as other kids.
This tech could mean that my children don’t have to face the same decision as their elders having to decide between dangerous and expensive surgeries for an unknown risk or living your life in fear.
I can’t tell you how huge this is for us if regular checkups would be enough to say, yes go live and play, enjoy your life. Things look good.
While I have no doubt that imaging-based neural nets can outperform a lot of the cardiac risk scores in clinical use today (they are usually pretty rudimentary), I would be fairly cautious about this article. Let's say one of the authors has a pretty long record of making extraordinary claims backed by very little evidence; managing to impress the medical community while at the same time being regarded as an impostor in the computer science community. Of course I'm just a random stranger on the internet, so I might just be a jealous competitor ;)
In health care domains we desperately want to believe providers know precisely what they're doing. It's particularly vexing to find out that "facts" are surrounded by an intrinsic level of uncertainty. IOW whatever action a practitioner deems best in a given situation is a judgement call grounded in experience. Sure, that's bolstered by underlying science, systematic investigations and published reports, but still there's no escaping the probabilistic nature of knowledge in the fields of medicine.
The provider you want to have is one who strives to do what's possible up to the limits of techniques and technology, and above all seeks to learn from every patient encounter.
I need to choose words better - rather than "dark arts" I probably should have said "tacit knowledge". I certainly didn't mean to imply that arrhythmias are not well studied, rather, my hope is that we can use AI methods to do better science.
side note: I know EEs who claim antenna design is a black art :-) It also is well studied and yet according to some, it appears magical at times.
Why isn't the method ("algorithm") made clear and presented to actual human experts so they can incorporate it in their diagnostic routines?
I'm sorry to be cynical, but all I've seen so far from "medical AI" is spurious correlational relationships in data, sometimes even in noise, presented as a causal deterministic predictions.
If you found a new way of predicting cardiac arrest, that's great. Let us know how.
https://wccftech.com/apple-watch-saves-life-of-a-70-year-old...
Additionally EKG’s puts leads all over your chest to get precise electrical measurements. A watch on your wrist isn’t going to be comparable, esp with things like sudden cardiac death. This article takes it even further as you need detailed imaging to really save these types of patients. Since we don’t yet know many of the underlying genetic variants or causes of sudden cardiac death. It’s more common than people realize and often goes undiagnosed or misdiagnosis because it’s hard to detect until too late.
Source: Multiple people in my family have died from sudden cardiac death. All with perfectly normal EKG’s before they passed.