Automatic quantification of myocardial perfusion stress-rest change: a new measure of ischemia

In myocardial perfusion SPECT (MPS), ischemia is typically quantified as the difference between stress and rest defect sizes obtained by separate comparisons with stress and rest normal limits. Such an approach is not optimal because images are not compared directly with each other and a complex set of stress and rest normal limits is required.

We developed a fully automatic technique to quantify stress-rest change. We applied it to 204 patients whose SPECT images were acquired using a same-day dual-isotope (99m)Tc/(201)Tl protocol and on whom coronary angiography had been performed. A 10-parameter registration of rest and stress images was performed by an iterative search of best translational, rotational, scaling, and optimal stress-rest count normalization parameters. Identical stress-rest 3-dimensional left ventricle (LV) contours were automatically derived from stress images. Integrated deficit counts (normalized rest-stress) within the LV volume were derived from registered image pairs. A global measure of ischemia (ISCH) was calculated as the ratio of the total deficit stress LV counts to the total rest LV counts.

Registration and derivation of quantitative measures were fully automatic. The average processing time was <40 s on a 2-GHz processor. When compared for prediction of stenosis, the area under the receiver operating characteristic curve (0.88 +/- 0.03) was significantly better for ISCH than that obtained by existing quantitative approaches, which use reference databases (0.80-0.82 +/- 0.03). The normalized stress-rest change could be visualized and localized directly on raw patient images using overlay display.

Automatic stress-rest MPS image registration allows a direct estimation of ischemia from SPECT that does not require comparisons with normal limits.