Alan Maniet

Computer-assisted myocardial thickening analysis of gated MIBI SPECT images.

RATIONALE AND OBJECTIVES The purpose of this study is to use the relationship between the partial volume effect, count density, matrix size, and reconstruction filter in gated single photon emission computer tomography (SPECT) to recover myocardial thickness, and to validate measurements generated using a computer-assisted automatic contour method with a phantom model and with thickness changes measured by echocardiography. METHODS Regional myocardial contour was defined automatically using shape constraints, gray level thresholding, and a gradient method applied to gated technetium-99m (99mTc)-methoxy-isobutyl-isonitile (MIBI) SPECT images. A heart phantom with wall thickness ranging from 0.8 cm to 1.3 cm was constructed. SPECT images were acquired and reconstructed using different matrix sizes, count densities, and filters. Wall thickness determinations derived from the reconstructions made with each combination of matrix size, type of filter, and cutoff frequency were correlated with the measured thicknesses. The best combination was applied to the gated MIBI SPECT images of 55 patients who also had echocardiography and coronary arteriography. Count density measurements were made across each regional myocardial segment to evaluate wall thickness at end-diastole and end-systole. Systolic wall thickening measurements made with this automated computer-assisted contour technique were compared with echocardiographic measurements made from segments with normal perfusion and from regions containing infarcted tissue. RESULTS Different reconstruction filters and cutoff frequencies affected the accuracy of measurements of myocardial wall thickness determined from gated myocardial SPECT images. A matrix size of 64 x 64, with a minimum of 10 counts/pixel/frame, and use of a Hanning filter with 0.5 cyc/cm cutoff frequency gave the best combination for myocardial thickness determination and spatial resolution. Application of these factors to the phantom yielded results that correlated very well with the thickness measurements (r = 0.986, P < 0.001). Application of the technique to the clinical SPECT studies yielded measurements of myocardial wall thickening that were not significantly different from that determined by echocardiography. CONCLUSION Computer-assisted contour analysis of gated SPECT images enables accurate determination of regional wall thickening using the count density changes within each myocardial segment.