Machine Learning in the Evaluation of Myocardial Ischemia Through Nuclear Cardiology

Abstract

To summarize the advances achieved in the detection and characterization of myocardial ischemia and prediction of related outcomes through machine learning (ML)-based artificial intelligence (AI) workflows in both single-photon emission computed tomography (SPECT) and positron emission tomography (PET). In the field of cardiology, the implementation of ML algorithms has recently gravitated around image processing for characterization, diagnostic, and prognostic purposes. Nuclear cardiology represents a particular niche for AI as it deals with complex images of semi-quantitative and quantitative nature acquired with SPECT and PET. AI is revolutionizing clinical research. Since the recent convergence of powerful ML algorithms and increasing computational power, the study of very large datasets has demonstrated that clinical classification and prediction can be optimized by exploring very high-dimensional non-linear patterns. In the evaluation of myocardial ischemia, ML is optimizing the recognition of perfusion abnormalities beyond traditional measures and refining prediction of adverse cardiovascular events at the individual-patient level.