Myocardial Perfusion Imaging is widely used to evaluate left ventricular perfusion in patients with ischemic heart disease. Semi-quantitative scores, such as Total Perfusion Deficit (TPD), are frequently used for the assessment of extent and severity of disease. TPD scores are based on a frequentist analysis of tracer distribution in perfusion polar maps, which are 2D polar representations of radiopharmaceutical tracer uptake in the left ventricular cardiac wall. In this paper, we propose a novel quantification method for MPI examinations. Our method approaches the problem from an inverse perspective to TPD, allowing the assessment of the severity and extension of abnormal perfusion from the synergetic contribution of all the left ventricular tracer uptake in a polar map. We analytically demonstrate that TPD can be viewed as a specific instance of a broader approach. Furthermore, we explore how to establish the quantitative analysis in a general approach. To test our new method, we leveraged a standard deep learning architecture for the prediction of abnormal perfusion and used gradient-based class activation maps as a measure of the abnormal perfusion territory, which facilitated the quantification of both the severity and extent of the perfusion deficit. In the absence of ground-truth data such as survival rates or major adverse cardiovascular events (MACE) criteria, our results demonstrate both statistical and visual consistency with the conventional TPD.
Guo et al. (Sun,) studied this question.