Correspondence between gated SPECT and cardiac magnetic resonance quantified myocardial wall thickening

Abstract

Left ventricular (LV) wall thickening (WT) assessed from myocardial perfusion (MP) gated SPECT data has been reported to be a marker of functional recovery following myocardial damage. However, the accuracy of WT measurements obtained in the clinical setting rarely has been validated against an independent quantitative reference standard. The purpose of this investigation was to assess the degree to which quantified MP WT agrees with cardiac magnetic resonance (CMR) WT measurements, and to determine whether quantitation is as accurate as visual analysis in detecting abnormal regional WT. MP and ECG-gated True-FISP CMR data were analyzed for 20 patients evaluated after myocardial infarction (age 60 ± 11 years; 95% males). An experienced observer visually graded MP WT on a 5-point scale while viewing MP cines. MP WT was quantified using "Emory Cardiac Toolbox" (ECTb) algorithms. MP algorithms isolated myocardial counts and generated polar maps of WT. CMR data were analyzed by Medis "MASS" software. Manually drawn endocardial and epicardial contours were used to compute WT on CMR. CMR data also were processed for 10 age-matched normal volunteers to define the CMR WT threshold of abnormality. All computations were sampled into conventional 17 ACC/AHA LV wall segments. Receiver operating characteristics (ROC) curve analysis provided discrimination thresholds for optimal accuracy, which subsequently were used to dichotomize the MP methods. WT abnormalities also were assessed for the 3 major arterial territories, and for total numbers of abnormal segments per patient. 25% of all segments had abnormally low WT by CMR. While MP quantitation underestimated CMR WT values for segments with normal WT (26 ± 13% vs. 56 ± 28%, P < 0.0001), measurements were similar for segments with abnormal WT (4 ± 12% vs. 5 ± 9%, P = 0.45). On a segment-by-segment basis, detection of abnormal WT was more accurate by quantitative than visual analysis both for continuous variables (ROC area = 88 ± 2% vs. 80 ± 3%, P < 0.0001) and for dichotomized methods (83% vs. 76%, P = 0.04). Agreement of MP versus CMR for discriminating segments with normal from abnormal WT was significantly better for quantitative than visual analysis (κ = 0.59 vs. 0.40, P < 0.0001), with strongest agreement for left anterior descending artery territories (κ = 0.72). Total numbers of segments with abnormal WT per patient demonstrated significant correlation with CMR (r = 0.83, P < 0.0001). MP quantified LV ejection fractions and volumes also correlated well with CMR (r = 0.87 and 0.90, respectively). Quantified MP WT measurements correlated significantly with CMR values, and discriminated segments with abnormal WT from segments with normal WT more accurately than visual analysis. Therefore, quantification should be performed when analyzing regional WT by scintigraphy.