Automated assessment of myocardial SPECT perfusion images: a comparison of case based reasoning, polar map analysis, and visual interpretation

This study compares the automated assessment of myocardial SPECT perfusion scintigrams by case based reasoning (CBR) with polar map analysis and visual interpretation, CBR algorithms match a given image to cases stored in a library of normal and abnormal images and identify one or more images that are similar to the image to be analyzed. The authors hypothesized that such an approach will show increased accuracy as its analysis not only uses data from normal subjects as is the case for polar map analysis but also employs abnormal image data. The authors tested this hypothesis in 240 pts (182 M, 58 F, age: 61±12 yrs) that underwent coronary angiography and perfusion scintigraphy within 6±11 days. By ROC analysis, the diagnostic accuracy of CBR was not different from polar map analysis and visual interpretation (p=n.s.). Sensitivity at 50% and 80% specificity were 90% and 67% for CBR, 88% and 65% for polar map analysis, and 91% and 74% for visual interpretation of scintigrams by an experienced observer. Thus, automated interpretation of myocardial SPECT perfusion images by case based reasoning shows similar diagnostic accuracy as PM analysis and visual interpretation when compared to coronary angiography