Getting your real-time heart rhythm monitored for 10 days now requires an ECG technician to review and notify your doctor of a critical arrhythmia. A new study showed A.I. did better than technicians (98.6% vs 80.3%) for identifying these abnormalities nature.com/articles/s41... - ThreadSky

erictopol.bsky.social • 12 days ago

Getting your real-time heart rhythm monitored for 10 days now requires an ECG technician to review and notify your doctor of a critical arrhythmia. A new study showed A.I. did better than technicians (98.6% vs 80.3%) for identifying these abnormalities
https://nature.com/articles/s41591-025-03516-x

Comments

craziedaizee.bsky.social•12 days ago

Hoping someday for open-source collective
with their own AI analysis where you can upload
ECG files made by KardiaMobile etc for diagnostic

FDA has to approve analysis and that's holding back hardware developers?

So split the function, let us DIY analysis
but that's where they make money on upsell

opus3.bsky.social•12 days ago

Old ICU/CCU nurse here… I will always prefer humans with lots of experience.

crystalveronie.bsky.social•12 days ago

This implies that ECG technicians do not already use tech to collect data and analyze it. The message is that AI can replace the technician. This is ridiculous. The tech is trained to the tech. You can add AI to current tech used, but this just means that technicians will need advanced training.

verityplayer.bsky.social•12 days ago

I dread to think what an AI would make of my ECG - probably say I've died.

tccschreiner.bsky.social•12 days ago

See, this is the kind of thing AI is great for--to analyze data and let a human know that something here is worth verifying. We don't need SmarterChild with Google access and steals things made by humans.

bkmnus.bsky.social•12 days ago

Reviewing hours/days worth of info, when the technician likely has many other duties, makes missing events more likely. I doubt there are many institutions that have a person whose only job duty is to review monitoring data.

nclud.bsky.social•12 days ago

Building classifier systems is great! It’s what sifting through massive amounts of data is useful for. The human mind is also a fantastic pattern recognition machine. iirc, the gen-ai picked image generators started up initially by taking the image classifier neural nets and running them backwards.

reginaedelman.bsky.social•12 days ago

and AI is made by humans too.

ber-droid.bsky.social•11 days ago

Truly amazing results. As always, the gain in sensitivity comes at a cost: the reduced specificity, i.e., more false positives. This needs to be considered to evaluate the overall impact of such tools.

reginaedelman.bsky.social•12 days ago

Who makes AI is OI=organic intelligence
Maybe the technician was not doing his duty correct.

lucidrains.bsky.social•12 days ago

ah, hybrid cnn + attention

adkraine.bsky.social•12 days ago

Well, I hope that the arrhythmia recognition has improved.
13 years in CCU with “smart” monitors still required a trained nurse or technician to verify lethal ones from artifact. And respond appropriately.

We had so many false alarms that often they were disabled.

adkraine.bsky.social•12 days ago

But we never defibrillated anyone brushing their teeth.

Comments

Posting Rules

Reply