Featured Publications

Publications En Vedette

DeepCORO-CLIP

Sarra Harrabi, Yichen Wu, Omar Majwal, Gabriel Asslo, George Azar, Geoffrey H. Tison, Minhaj Ansari, Milos Vukadinovic, David Ouyang, Joshua P. Barrios, Jacques Delfrate, Robert Avram

Les cardiologues interprètent une coronarographie en examinant plusieurs vues des mêmes artères. DeepCORO-CLIP réunit ces vues dans un modèle entraîné sur 203 808 vidéos d’angiographie associées à des comptes rendus de cathétérisme provenant de 28 117 patients. Il évalue les sténoses, les occlusions totales chroniques, les thrombus et les calcifications dans les différents segments coronariens. Le même encodeur préentraîné peut être adapté pour estimer la fraction d’éjection ventriculaire gauche et prédire des événements cardiovasculaires à partir de l’angiographie seule. L’étude comprend une validation externe à l’UCSF, une comparaison avec des mesures réalisées par un laboratoire central indépendant et une évaluation en aveugle, par des cardiologues, de comptes rendus générés automatiquement.

October 13, 2026

Rohan Banerjee, Juliette Beaulieu, Nicolas Dostie, Blandine Mondésert, Ram Ahuja, Alexis Nolin-Lapalme MD-PhD, Gilbert Jabbour, Shreya Shree Srikanth, Guillaume Marquis-Gravel, Olivier Tastet, Achille Sowa, Julia Cadrin-Tourigny MD-PhD, Jacques Delfrate MSc, Robert Avram MD-MSc

Most ECG-AI systems are built for a single fixed task and return predefined labels rather than the reports or patient-specific answers clinicians need. DeepECG-Tok converts raw 12-lead ECG waveforms into discrete tokens that can condition a medical language model, allowing one instruction-tuned system to interpret ECGs, generate structured reports and predict endpoints including left ventricular function, structural heart disease, incident atrial fibrillation, acute coronary occlusion and the likely culprit artery. Trained on more than 7.27 million question-answer pairs and evaluated across four external cohorts, the model retained strong diagnostic performance and produced free-text reports rated comparably to cardiologist references.

DeepRV central figure

Fatima Zahra Fawzi, MSc, Istok Menkovic, MD, Nicolas Dostie, MD, Maxime Tremblay-Gravel, MD, MSc, Marie-Claude Parent, MD, Gabriel Assolo, MD, Jean-François Tanguay, MD, PhD, Guillaume Marquis-Gravel, MD, MSc, Minhaj Ansari, MSc, Joshua P. Barrios, PhD, Geoffrey H. Tison, MD, MPH, Jacques Delfrate, MSc, Robert Avram, MD

Right ventricular systolic function (RVSF) is a critical determinant of cardiovascular outcomes, yet rapid assessment during coronary angiography remains challenging when prior imaging is unavailable. DeepRV is an open-weight video-based deep learning model that predicts RVSF directly from routine coronary angiograms, enabling real-time RV dysfunction screening in the cath lab.

Alexis Nolin-Lapalme, Achille Sowa, Jacques Delfrate, Olivier Tastet, Denis Corbin, Merve Kulbay, Derman Ozdemir, Marie-Jeanne Noël, François-Christophe Marois-Blanchet, François Harvey, Surbhi Sharma, Minhaj Ansari, I-Min Chiu, Valentina D'souza, Sam F. Friedman, Michaël Chassé, Brian J. Potter, Jonathan Afilalo, Pierre Adil Elias, Gilbert Jabbour, Mourad Bahani, Marie-Pierre Dubé, Patrick M. Boyle, Neal A. Chatterjee, Joshua Barrios, Geoffrey H. Tison, David Ouyang, Mahnaz Maddah, Shaan Khurshid, Julia Cadrin-Tourigny, Rafik Tadros, Julie Hussin, Robert Avram

The 12-lead electrocardiogram (ECG) remains a cornerstone of cardiovascular diagnostics, with over 300 million performed annually worldwide. Existing AI solutions for automated interpretation often lack generalizability, remain closed source, and depend on supervised learning requiring extensive labelled datasets.

January 23, 2026

Pascal Thériault-Lauzier, M.D., Ph.D. , Nils Perrin, M.D., M.Sc., Arash Sarshoghi, B.Sc., Stella Ly, Michael Knafo, Aurélie Grandchamp , Yan Xu Rong, Marie-Gabrielle Lessard, R.N., Denis Corbin, B. Eng, M.Sc., Bahareh Taji, B. Eng., Ph.D., Olivier Tastet, M.Sc., Aun Yeong Chong, M.D., Michael Froeschl, M.D., M.S., Alexander Dick, M.D., Marino Labinaz, M.D., Christopher Glover, M.D., Michel Le May, M.D., Zeeshan Ahmed, M.D., Omar Abdel-Razek, M.D., M.S., ...Pietro Di Santo, M.D., Ph.D., David Messika-Zeitoun, M.D., Ph.D., Geoffrey H. Tison, M.D., M.P.H., Juan Russo, M.D., Guillaume Marquis-Gravel, M.D., M.Sc., Jean-François Tanguay, M.D., Richard Gallo, M.D., Quoc Hung Ly, M.D., Anita Asgar, M.D., Serge Doucet, M.D., Gilbert Gosselin, M.D., Jean Grégoire, M.D., Reda Ibrahim, M.D., Philippe L. L’allier, M.D., Mohamed Nosair, M.D., Derek Y.F. So, M.D., M.Sc., and Robert Avram, M.D., M.S.

Accurate left ventricular ejectionfraction (LVEF) assessment is crucial in acute coronary syndrome (ACS)management but traditional methods like echocardiography (TTE) andventriculography have limitations such as risks, contrast exposure, andlogistical delays.The CathEF study introduced adeep-learning algorithm integrated with the PACS-AI (www.pacsai.co) platform,providing real-time LVEF estimates directly from routine coronary angiographyvideos, without extra catheterization or contrast.

Gilbert Jabbour, Alexis Nolin-Lapalme, Olivier Tastet, Denis Corbin, Paloma Jordà, Achille Sowa, Jacques Delfrate, David Busseuil, Julie Hussin, Marie-Pierre Dubé, Jean-Claude Tardif, Léna Rivard, Laurent Macle, Julia Cadrin-Tourigny, Paul Khairy, Robert Avram*, Rafik Tadros*

Atrial fibrillation is the most common sustained cardiac arrhythmia in adults and is associated with an increased risk of stroke, heart failure, cognitive decline, hospitalisations, and death. Deep learning applied to electrocardiograms (ECG-AI) is an emerging approach for predicting atrial fibrillation or flutter (AF). We introduce an open-weights ECG-AI model developed at the Montreal Heart Institute (MHI) and externally validated using MIMIC-IV dataset, comparing its performance with clinical and AF polygenic scores (PGS).

August 30, 2024

Élodie Labrecque Langlais, Denis Corbin, Olivier Tastet, Ahmad Hayek, Gemina Doolub, Sebastián Mrad, Jean-Claude Tardif, Jean-François Tanguay, Guillaume Marquis-Gravel, Geoffrey H. Tison, Samuel Kadoury, William Le, Richard Gallo, Frederic Lesage & Robert Avram

Coronary artery disease involves the narrowing of heart blood vessels of the heart, also known as. stenoses, and stands as a major global cause of mortality, requiring accurate evaluation for effective treatment. The existing gold standard for assessing stenosis severity, which relies on visual interpretation of coronary angiography videos by cardiologists, shows significant variability, highlighting the necessity for standardized, objective methods. AI-based algorithms have been developed to overcome this issue, but these methods have inherent limitations which are addressed with DeepCoro.

Robert Avram, Jeffrey E. Olgin, Zeeshan Ahmed, Louis Verreault-Julien, Alvin Wan, Joshua Barrios, Sean Abreau, Derek Wan, Joseph E. Gonzalez, Jean-Claude Tardif, Derek Y. So, Krishan Soni & Geoffrey H. Tison

Coronary heart disease (CHD) is the leading cause of adult death worldwide. The minimally-invasive procedure called coronary angiography provides the primary gateway for nearly all CHD diagnosis and clinical management decisions, as well as stent-based therapy for CHD.

October 22, 2023

Robert Avram, Joshua P Barrios, Sean Abreau, Cheng Yee Goh, Zeeshan Ahmed, Kevin Chung, Derek Y So, Jeffrey E Olgin, Geoffrey H Tison

A team of researchers at UCSF developed a technology called CathEF that uses artificial intelligence (AI) to assess how well the heart pumps blood during a common heart test.