The Innovation Forum, presented by the ACC Health Care Innovation Section, is a quarterly virtual event hosted by Regina Druz, MD, FACC, and David Cho, MD, MBA. January's Innovation Forum doubled as the inaugural "AI in Cardiology" journal club organized by the Advanced Health Care and Analytics Work Group. Titled "Computer Vision: A Clinician's Third Eye," the event focused on the research, education, and real-life application of AI in Cardiology through an excellent session guided by a distinguished group of domain experts in both academia and the industry. Addressing the research and education components was a presentation by Partho Sengupta, MD, FACC, followed by a thought-provoking discussion on the deployment of AI led by an expert panel, including Anthony Chang, MD; Matthew Lungren, MD, MPH; and Paul Lee, MD. Jai Nahar, MD, and Francisco Lopez- Jimenez, MD, FACC, moderated the session.

In his presentation, Sengupta guided us through the importance of machine vision and machine learning in cardiology. The article for discussion was his state-of-the-art review on "Proposed Requirement for Cardiovascular Imaging-Related Machine Learning Evaluation (PRIME): A Checklist," in JACC: Cardiovascular Imaging. Put together by a multidisciplinary panel of machine learning experts, clinicians and statisticians, and led by Sengupta, the checklist aimed to create a uniform, standardized reporting of machine learning investigation which would aid authors, reviewers and readers of such work. It goes on to highlight the seven critical steps in a machine learning pipeline, namely, defining the project goal, data cleaning and preparation, data description and visualization, model building, interpretation and reporting, reproducibility, and description of limitations and alternatives. Furthermore, he took us through the landmark article in Circulation on fully automated echocardiogram interpretation by Zhang et al. to appreciate the seven items in action.

The second half of the session addressed the real-life application of AI in medicine and cardiology. The panelists addressed questions related to training clinical cardiologists in AI, challenges around the availability of accurately labeled data, and collaborative and educational opportunities in various institutions. An active area of research is the integration of "pixel-based" data with clinical and genomic data. Lungren's recent work on pulmonary embolism is a step in that direction. This is especially relevant in cardiology, where the integration of ECG, echo, angiogram and clinical data would make these algorithms more robust. To that end, Chang proposed the idea for an open-source "cardiomic" database, making all these data available on one common platform, ready to be used for research.

Another crucial point that was discussed was the role of explainable AI. Algorithms need not only be accurate but explainable as well so that the science behind the predictions can be leveraged to develop early therapeutic interventions. That said, explainability might not be very important in the prospective use of the model as long as it was factored in during the development. Furthermore, Lee shared the inspiring story of developing his iPhone application, reflective of the indomitable spirit of a cardiologist.

In conclusion, the inaugural AI in Cardiology journal club was an absolute treat with plenty of great teaching points. As someone very interested in this space, I cannot wait for the next one!

A recording of the Jan. 28 Innovation Forum, Computer Vision: A Clinician's Third Eye is available here.