As a Cardiologist with expertise in AI, I definitely see a whole lot of areas where AI can improve the practice of cardiology.  But my excitement is less about AI adding more accuracy to the reading of the variety of the imaging studies performed in the assessment of the cardiac patients, and more about adding intelligent automation to the practice of cardiology. What does that mean? Cardiovascular disease is the number one killer in America and number two in the world. Much of this is due to lifestyle factors, lack of adequate early screening and intervention, cumbersome diagnosis and scarcity of specialists.

We have good interventions to prevent, slow, or treat heart disease. We just need to identify those at risk and intervene sooner. If AI algorithms can allow us to use the modern technologies such as cell phones, telehealth, intelligent bots, etc to collect data, diagnose, and intervene in a more streamlined fashion, we can lower the burden of cardiovascular disease significantly. That means less suffering and lower costs. As such, careful thinking through how AI algorithms can be applied to data, that can increasingly come from unconventional channels, to diagnose disease will be required to ensure these solutions gain traction and can scale up.

Some of the early efforts in using AI in the practice of cardiology involves augmenting the cardiologists in interpreting some of the commonly performed tests such as electrocadiogram (ECG) and echocardiogram. Currently, when an ECG is performed, a rules-based algorithm provides an initial interpretation that is often ignored by the cardiologist (very rudimentary findings and not robust enough.) Early evidence suggests that deep neural network algorithms can vastly improve over this reading. There’s emerging evidence that AI algorithms can be effective in using data collected from cell phone-based rhythm detection systems to diagnose arrhythmias.