Novel indices representing heterogeneous distributions of myocardial perfusion imaging

Abstract

Abstract Introduction Heterogeneous distribution in myocardial perfusion images (MPI) obtained by scintigraphy is often observed in cardiac diseases with normal myocardial perfusion. However, quantitative assessments of such heterogeneity have not been established. We hypothesized that the heterogeneity in MPI can be quantitatively evaluated through histogram analysis, calculating the standard deviation (SD), the 95% bandwidth (BW95%), and entropy. Methods We examined resting 99mTc-MIBI images in 20 healthy subjects and 29 patients with cardiac disease who had none or very-mild reduced myocardial perfusion evaluated as a low summed rest score (0 to 4, the range of the studied healthy subjects). Two nuclear medicine specialists blindly divided them into two groups: non-heterogeneity or heterogeneity group, based solely on their visual assessments of heterogeneity on splash and polar maps generated from single-photon emission computed tomography (SPECT) images. The %uptake was determined by dividing the tracer count of each pixel by the tracer count of the pixel with the highest value in the LV myocardium. SD, BW95%, and entropy from histogram patterns were analyzed from the polar map data array of each %uptake. We investigated whether heterogeneity could be assessed using SD, BW95, and entropy in two groups classified by visual assessments. Additionally, we evaluated the area under the curve (AUC) to identify heterogeneity in the receiver operating characteristic curve analysis. Results Based solely on visual assessments, 11 (22%) and 38 (78%) cases were classified into the non-heterogeneity and heterogeneity groups, respectively. The non-heterogeneity group consisted of only healthy subjects, and all patients with cardiac disease were classified into the heterogeneity group. The cases in the heterogeneity group had significantly higher values of heterogeneity indices (SD, BW95%, and entropy) in %uptake than those in the non-heterogeneity group (p < 0.05 for all). The AUCs of the heterogeneity indices were sufficiently high (AUCs > 0.90 for all) in distinguishing cases with visually heterogeneous distribution or patients with cardiac disease. Conclusions Heterogeneity in MPI can be evaluated using SD, BW95%, and entropy through histogram analysis. These novel indices may help identify patients with subtle myocardial changes, even in images that show preserved perfusion (345/350).