Diagnostic accuracy of an artificial neural network compared with statistical quantitation of myocardial perfusion images: a Japanese multicenter study

1,365 patients (1,001 training, 364 validation) with confirmed or suspected coronary artery disease who underwent stress/rest myocardial perfusion imaging. Validation cohort: mean age 71, 73% male, from 9 hospitals in Japan. Excluded: LBBB, severe cardiomyopathy, moderate/severe valvular heart diseases.

Artificial neural network (ANN) based diagnostic system for interpreting myocardial perfusion images

Conventional statistical quantitation using summed stress/rest/difference scores (SSS/SRS/SDS)

Diagnostic accuracy (Area under the ROC curve, AUC) for identifying stress defects, rest defects, and stress-induced ischemia, compared to expert interpretation as the gold standardsurrogate