
Future chief resident @NYULangone. AI-ECG research 🫀 MD @HarvardMed, BA @Harvard. Opinions my own.
Manhattan, NY
Joined November 2022
- Tweets27
- Following794
- Followers193
- Likes106
Constantine Tarabanis, MD retweeted
Recent trials suggest that digitalis may benefit selected symptomatic #HFrEF patients, but modest effects, safety, and fragility warrant caution @shaan_khurshid @DrCardioCon @GivertzMichael ahajrnls.org/47emhCF
Constantine Tarabanis, MD retweeted
Our home-grown AF from Sinus AI prediction model published today in #EHJ digital health.
tldr: We reach AUC 0.9 that holds on external validation across continents!💥
@ehj_ed @nyuniversity @nyulangone @DrCardioCon
Read here:
academic.oup.com/ehjdh/artic…
Constantine Tarabanis, MD retweeted
Excited to share our new study! @JAMAPsych @heffron_sean @NYUCVDPrevent @YaleIMed @TradIMYale
Meta-Analysis looking at Suicide and Self-harm events in patient with Obesity or Diabetes taking GLP1-RA
Link:
jamanetwork.com/journals/jam…
Constantine Tarabanis, MD retweeted
New paper out today!
TLDR:
We use special self supervised learning architecture called VICReg, invented by @ylecun to solve the problem of insufficient labeled data
We show that real prevalence of "rare" diseases may be much higher than perceived 😱
nature.com/articles/s41598-0…
Constantine Tarabanis, MD retweeted
AI provides a universal framework that leverages data and compute at scale to uncover higher-order patterns
Today, @arcinstitute in collaboration with @nvidia releases Evo 2—a fully open source biological foundation model trained on genomes spanning the entire tree of life 🧵
Constantine Tarabanis, MD retweeted
A multimodal foundation model for medical images from 15 million image-text pairings ai.nejm.org/doi/full/10.1056… @NEJM_AI open-source, open-access paper
@hoifungpoon @MSFTResearch
Constantine Tarabanis, MD retweeted
A randomized trial of tailored A.I.-guided ablation of atrial fibrillation vs standard-of-care, anatomical ablation demonstrated superiority of AI guidance
nature.com/articles/s41591-0…
@NatureMedicine
Constantine Tarabanis, MD retweeted
1/n We are excited to announce EchoPrime – the first echocardiography AI model capable of evaluating a full transthoracic echocardiogram study, identify the most relevant videos, and produce a comprehensive interpretation!
Great work lead by @milos_ai, EchoPrime is the largest and most complex #echofirst AI model yet, trained on 10x the data of existing models.
Constantine Tarabanis, MD retweeted
Professor Charged for Operating Multimillion-Dollar Grant Fraud Scheme
Wang allegedly engaged in a scheme to fabricate and falsify scientific data in grant applications made to the NIH on behalf of himself and the biopharmaceutical company.
$SAVA
justice.gov/opa/pr/professor…
Constantine Tarabanis, MD retweeted
Prospective RCT of AI-Enhanced ECG for STEMI Activation @NEJM_AI
ai.nejm.org/doi/full/10.1056…
Excellent trial building upon and complementing recent work in @NatureMedicine nature.com/articles/s41591-0…
Excellent editorial by @RobertAvramMD and @wfearonmd
ai.nejm.org/doi/full/10.1056…
AI in 🫀
🎉🎉 Our 2nd AI-ECG model dealing with the QT interval, this time for diLQTS ✅ safe prescribing in the outpatient setting.
Reproducible + Standardized nature of ECGs can make them a better input for medical AI models than poorly curated EMR tabular data 📊.
⚡️Just out @JACCJournals⚡️
Prediction of 💊 induced QT prolongation with #AI
💥QTNet - 🆕 architecture for fusing ECG with EHR data
⏱️-dependent prediction from day 0- 180 & external validation
💥day 2 AUC > 0.9
sciencedirect.com/science/ar…
💫@neiljethani
💫@DrCardioCon
💫HaoZhang
1
454
Constantine Tarabanis, MD retweeted
Delighted to share ✨Med-Gemini✨ - our new family of multimodal models for medicine unlocking new possibilities for health - arxiv.org/pdf/2404.18416
More accurate multimodal conversations about medical images🩻, surgical videos📽️, genomics🧬, ultra-long health records📚, ECGs🫀 & more with state-of-art performance across multiple benchmarks
More accurate, up-to-date answers to medical questions with advanced reasoning and intelligent use of web-search
Long-context abilities. Summaries or referral letters from long health records, analyses of dozens of long research PDFs & more (1/6)
Constantine Tarabanis, MD retweeted
Excited to share our #echofirst foundation model is now published in @NatureMedicine!
Trained on 1M @CedarsSinai A4C echo-report pairs, EchoCLIP has strong performance on wide range of cardiac diagnostic benchmarks, including LVEF in @StanfordMed videos (MAE 7.1%).
What happens when you use >1 million echo video clips with expert interpretation to train an #AI?
nature.com/articles/s41591-0…
Nice advance by @David_Ouyang and colleagues @CedarsSinai @CedarsSinaiMed @NatureMedicine
AI-ECG RCT
Alert: pts with ↑ mortality risk
1o endpoint: ↓ all cause mortality
Consent: physicians only
Generalizability limitations, but supporting the value of AI models based on standardized & reproducible inputs (ECG>EMR tabular data) nature.com/articles/s41591-0…
216
Constantine Tarabanis, MD retweeted
Be skeptical of the findings in single-centre RCTs of critically ill patients published in high-impact journals claiming a mortality benefit.
Only 1 out of 16 RCTs were replicated in a multi-centred RCT. ccforum.biomedcentral.com/ar…
@First10EM @KirstyChallen @drjohnm @adamcifu
Constantine Tarabanis, MD retweeted
🚨Stunning!🚨
REVERSAL in the decline of heart failure ☠️ in the US 1999-2021
1999-2005 ⬇️ ☠️
2005-2012 ⬇️ ☠️
2012-2019 ⬆️ ☠️
2019-2021 ⬆️⬆️☠️
Age-adjusted HF related ☠️ rates higher in 2021 vs 1999!
WT😱🆘
jamanetwork.com/journals/jam…
@JAMACardio @FudimMarat
Constantine Tarabanis, MD retweeted
A great weekend welcoming the new fellows to Brigham Heart! We cannot wait for you all to join us beginning in July! Welcome to Brigham Fellows 🫀🫀
Constantine Tarabanis, MD retweeted
#AI for rare arrhythmia disorders.
Inspired by the work of @ylecun on Self-Supervised Learning, we adapted VICReg, an Energy Based Model to learn deep #EKG representations.
Pre-training with SSL is a promising strategy for rare disease detection.
@nyulangone @nyuniversity
👇
Brugada ECG detection with self-supervised VICReg pre-training: a novel deep learning approach for rare cardiac diseases medrxiv.org/cgi/content/shor… #medRxiv





























