Gated SPECT offers improved interobserver agreement compared with echocardiography

Left ventricular ejection fraction (EF) is a powerful predictor of prognosis in coronary artery disease. The purpose of the present study was to measure interobserver differences for gated SPECT (GSPECT) software and echocardiography, and to compare these modalities regarding left ventricular volumes and EF.

Eighty-four patients scheduled for nuclear imaging underwent a 1-day GSPECT with Tc-99m-tetrofosmin. Images were processed by 2 raters who calculated volumes and EF using Cedar-Sinai quantitative gated-SPECT (QGS), Emory Cardiac Toolbox (ECT), and 4D-MSPECT of the University of Michigan. Echocardiographic volumes were measured by 2 raters. Interobserver reliability was assessed by intraclass correlation coefficient (ICC). Differences in volumes and EF between echocardiography and GSPECT were compared with t-tests.

ICC was 0.61 for echocardiography, 0.94 for QGS, 0.88 for ECT, and 0.91 for 4D-MSPECT (P < 0.0001 compared with echocardiography). For small ventricles (ESV ≤30 mL), ICC was 0.58 for echocardiography and 0.90 for QGS (P = 0.008 compared with echocardiography); 0.77 and 0.73 for ECT and 4D-MSPECT, respectively (P = ns). End-diastolic and end-systolic volumes were significantly larger with GSPECT than with echocardiography, also echocardiographic ejection fraction was significantly different from GSPECT.

There is better interobserver reliability in GSPECT as compared with echocardiography, and QGS seems more robust in this study especially when it comes to small ventricles.