A large multicenter study, published in eClinicalMedicine journal has shown that artificial intelligence (AI) can significantly improve the prediction of acute coronary syndrome (ACS) when applied to coronary CT angiography (CCTA) scans, a standard imaging test for suspected coronary artery disease (CAD).

Data was sourced from 4,436 patients across multiple institutions. Machine learning models were developed using both anatomical and plaque-related features visible on CCTA. The goal was to determine whether ML could enhance ACS risk prediction compared to traditional measures such as stenosis severity alone.

The machine learning model showed stronger performance than traditional stenosis-based assessment, with an AUC of 0.85 versus 0.72, indicating better categorization of patients who did and did not develop ACS. 
ML-based models provided greater accuracy in identifying patients who would later experience ACS, especially among those patients, where decision-making could be challenging. The model integrated subtle imaging features often overlooked in routine interpretation, which improved distinction and calibration.

These findings highlight the potential of AI to bridge the gap between imaging and personalized care. Timely and accurate identification of ACS risk could accelerate intervention, reduce unnecessary testing, and improve patient outcomes.

For now, an AI-based approach to identify ACS risk needs further research, preferably in prospective cohort trials. If widely implemented, this approach could streamline risk analysis, enabling cardiologists to prioritize the right patients for timely therapies and potentially save lives.