From detection to prevention: AI’s impact on cardiovascular care
Computing and medicine: a perfect match
Robert Avram was born in Romania and moved to Sherbrooke, Quebec with his family when he was 6. He developed a passion for computing and programming early on, spending his free time building websites and mobile apps. Because his parents encouraged him to pursue graduate studies in medicine, he redirected his focus to become a cardiologist. Unbeknownst to him, it was a career choice that would reconnect him with his childhood passion.
He then completed a postdoctoral fellowship in cardiology in San Francisco – the global tech capital – where he honed his vision. “I was conducting research in digital health and I noticed AI started to become widespread. It was a bit like science fiction for me! But I knew there was something there. I knew my expertise in cardiology could be valuable to the development of that particular field. So, I enrolled in a course on AI at Stanford University where we built an app designed to collect information about the health of patients and use AI to analyze the data. This type of technology can even be used to diagnose diabetes remotely. It opened my eyes to technology’s immense potential for extracting and interpreting data. My head was filled with ideas when I came back to Quebec,” he added.
The HeartWise.AI Lab: AI at the heart of breakthroughs
Dr. Avram has headed the Montreal Heart Institute’s HeartWise.AI Lab since July 2021. It’s a multidisciplinary research centre where nurses, engineers, students in bioinformatics, and physicians combine medical imaging, medical signal processing, and deep learning to create equitable and practical AI solutions. “We work together to develop solutions while always keeping in mind the reasons why we program these tools and their impact on people. Interdisciplinarity is the key to AI-powered tools. Engineers are able to extract data, physicians can interpret them and submit them for ethical review — a key step when it comes to health care and human lives,” he said.
Health care experts are mindful of the potential effects of their work and the vital importance of the new insights they've gained. That puts them in the best position to provide a framework for the paradigm shift currently taking place in medicine.
Electrocardiograms (ECGs) are the most common test in cardiology. They produce a graphic representation of the heart's electrical activity to enable experts to detect cardiac abnormalities. “Currently, the device used to interpret the results of an ECG makes mistakes 29% of the time. The information transmitted must always be validated by a cardiologist. This can either take several hours in an institution like ours or might never happen in a primary care setting. To this end, we built the DeepECG.ai platform that uses artificial intelligence to interpret ECGs and instantly produce reports with an accuracy rate of 98%. Moreover, thanks to algorithms able to analyze complex data in real-time, the tool can instantly detect structural heart disease, like heart failure and valve disease, with an 80% accuracy rate,” said Dr. Avram.
DeepECG.ai delivers quick results, identifies patients at risk, improves diagnoses, and facilitates triage. Dr. Avram believes it will lead to tailored strategies that can prevent serious complications such as strokes. This is the perfect example of how this technology will never replace medical professionals but rather complement their skills.
In fact, DeepECG is already being used at the MHI and will soon be made available to family doctors throughout the province. “We’re going to launch a study with frontline workers to see how we can make the platform more accessible. During a routine exam for instance, a doctor will be able to use AI to quickly detect a heart abnormality on the ECG. Right now, there’s a lot of interest for the tool because it could quickly identify which patients are at high risk, help with triage, and prevent the complications associated with structural heart disease. The key aspect of this study is to determine how we can consider new possibilities in a responsible way. We need to find out the real benefits of the knowledge we gain, looking beyond diagnostic accuracy to the tangible impact of using the tool. For instance, will it lead to a needless number of consultations with physicians? And will our health system be able to meet the demand? We need to seriously assess all the consequences that result from this type of technological breakthrough,” said Dr. Avram.
Another analytical tool developed by the HeartWise.AI Lab is called DeepCoro. It’s been adopted by hospitals both here and abroad and redefines how certain arterial problems are assessed during a coronary angiography. “I’m a hemodynamics specialist by training. Some might even call me a heart plumber! The wide variability in tests for arterial blockages means two physicians might interpret the same results differently, leading to different recommendations that can range from medication to heart surgery. We built DeepCoro to standardize this process through analytical models that analyze the heart's dynamics in angiographic videos over time. The goal is to provide a more consistent and accurate interpretation of arterial blockages and the heart’s overall strength. In other words, less variability than is currently happening when a human interprets this test. Think of it as a second pair of eyes for the physician. A tool that sees everything and helps them make better decisions regarding the most suitable treatment.
The PACS system is a universal computing infrastructure that enables hospitals, clinics, and other health care institutions to share, manage, and protect their patients’ X-ray images. Dr. Avram wanted to create AI tools that could be bundled together. In collaboration with the University of Ottawa, Mila, and Polytechnique Montréal, he came up with the idea of creating PACS-AI, an open-source software that connects to the PACS system and provides physicians with various analytical tools. “We wanted to create the equivalent of an ‘app store’ where a physician could choose the tools they need. For instance, an app they can use to analyze a patient's heart X-ray to determine the best brand of pacemaker for the situation. That might seem trivial, but it's something incredibly hard to determine and yet crucial in deciding what actions to take. The same goes for a slew of other tools that will be developed and bundled into this accessible and easy-to-use software,” said Dr. Avram.