Comparison of nineteen quantitative models for assessment of localized left ventricular wall motion abnormalities.

The usefulness of 19 models (8 of half-axis models, 3 radial models, and 8 segmental area models) for detection of local left ventricular wall motion abnormalities was analyzed in biplane cineventriculograms of 14 normals, 12 patients with anteroseptal, 10 with inferior, and 8 with posterolateral infarction. Extent of shortening of 22 half-axes and decrease of 24 areas were calculated in each projection. Mean values and one-sigma deviation were calculated for all axes in the group of normals and for those axes which pertained to segments of myocardial infarction in the group of coronary artery disease patients. Area model 3.5 yielded the largest number of significantly different areas between normal and infarcted ventricles. Sensitivity for recognition of myocardial infarction was calculated at 90% specificity by means of receiver operating characteristic curves. The greatest sensitivities were obtained with radial model 2.3, which yielded 83.3% true positive results for anteroseptal infarction, 90% for inferior infarction, and 87.5% for posterolateral infarctions. The best area model 3.5 yielded sensitivity between 67.5 and 80% only. Thus, in contrast to the literature, receiver operating characteristic analysis revealed radial model 2.3 to be most accurate in the assessment of local wall motion abnormalities.