🧵1/Today, we published a key milestone towards AI based cardiac screening in Nature. https://t.co/Lr3ymIrgz5 EchoNext outperformed cardiologists and found thousands of high-risk patients missed in routine care. We also made a version available to the world.

Excellent, entertaining and educational Year in Review for Clinical AI @NEJM_AI https://t.co/dsczPAAPyh Kudos @PierreEliasMD @Emily_Alsentzer

@Yuki_Sahashi presents EchoNet-Measurement, an open source pipeline for #echofirst annotation trained on 100x more data than prior models. 🚀YIA presentation 📖Simultaneous @JACCJournals pub 💻Open source code and weights

What makes a JACC worthy AI paper? We have updated the editor's page to provide clear guidance. Thanks to @hmkyale and @JACCJournals for making it a priority to clarify evolving standards!

🚨 New postdoc position in our lab @Berkeley_EECS! 🚨 (please retweet + share with relevant candidates) We seek applicants with experience in language modeling who are excited about high-impact applications in the health and social sciences! More info in thread 1/3

GPT-6 will have the intelligence to open a bakery instead of doing a PhD

Artificial Intelligence for Cardiovascular Care AI models can detect cardiovascular disorders including reduced ejection fraction, valvular heart disease, and cardiomyopathies from electrocardiograms with accuracy not previously achieved by human experts or technology All

What a great summary of our recent Nature paper in 2 mins!

How the electrocardiogram can identify previously undiagnosed structural heart disease with A.I. alone, better than cardiologists with A.I. support @Nature @PierreEliasMD @timpotsMD https://t.co/Jef9261QbC

Nature research paper: Detecting structural heart disease from electrocardiograms using AI https://t.co/xlwdplotaq

1/6 🏥 Structural heart disease (SHD) is common and often missed. What if a standard ECG could spot it—accurately, at scale, across hospitals, and patient types? Meet EchoNext, our AI model trained on 1.2M ECG-echo pairs. 👉 Published in Nature 🔗

Thank you to @timpotsMD for leading this work with me, and to the amazing CRADLE team for the years it took to build this.

We also released the largest public ECG+echo label dataset, and mini-model weights (out Friday). We hope this becomes the benchmark for AI & cardiology. Please go build a better future for us all.

This work shows AI can: ✅ Identify undiagnosed SHD currently missed ✅ Outperform experts ✅ Generalize across systems and pts ✅ Guide echo to those who need it ✅ Do it at scale We are now running the country's largest cardiac AI trial in 8 hospitals: https://t.co/pbFeFs0maT

🧵7/ Finally, we prospectively enrolled 100 patients in the DISCOVERY trial who had an ECG but no echo. EchoNext scores stratified SHD prevalence: 🔴 High-risk: 73% had newly diagnosed SHD 🟡 Moderate-risk: 28% 🟢 Low-risk: 6% We have found hundreds of new diagnoses since.

🧵4/ How does it compare to cardiologists? We ran a blinded reader study with 13 cardiologists reviewing 3200 ECGs w/ and w/o the AI score. Even with the AI score, cardiologists still underperformed compared to AI alone. I think this will improve as docs get used to new AI scores

🧵3/ We trained our model, EchoNext, on 1.2M ECG–echo pairs from 230K+ patients across 8 hospitals in NYC. Then we validated it at 4 health systems (NYP, Cedars-Sinai, UCSF, Montreal Heart Institute). We saw excellent performance across race, ethnicity, and setting.

🧵2/ Why SHD? We have mammograms & colonoscopies, but no equivalent screening for the most common cause of death:🫀diseases. Heart failure, valve disease, and PH are often diagnosed late. Echo is the gold standard, but is expensive limiting use. We need a scalable screen.

Thank you to @CNN and Clare Duffy for having me on to talk about our work using AI to advance medicine and where the field is headed for patients, clinicians, & the country. Spotify https://t.co/fCyaTgmH7M Apple Podcasts https://t.co/3OGxE7waBs YouTube https://t.co/DO9MFRBfkA

Check out the Stanford AI Index Report: @indexingai It's the best one-stop-shop for credible, current data on AI trends.