Improved accuracy of SPECT thallium-201 using wavelet prefiltering

Abstract

Background: Image noise significantly degrades the quality of nuclear cardiac images. Noise reduction using conventional Fourier techniques has become an essential part of reconstruction. Wavelet filtering has several theoretical advantages over Fourier based methods but has not been evaluated in clinical nuclear cardiology. Methods: We identified 695 patients referred for exercise stress testing using SPECT Tl-201 who either had less than a 5% probability of coronary disease or catheterization documented disease within 60 days of stress testing. Patients with documented MI, pathologic Q waves, LBBB, non-ischemic cardiomyopathy or prior bypass grafting were not included. There were 315 women and 380 men, 309 had coronary disease and 386 were low probability. Twenty men and 20 women were taken from the low probability group to create gender and filter specific normal databases. Quantitative SPECT analysis was performed by conventional Fourier techniques using a Butterworth filter with a cutoff of 0.4 of the Nyquist and an order of 8. Wavelet prefiltering was performed by transforming the each frame of the projection data with a 4 th order Coiflet filter then performing a soft threshold using 4.5x the standard deviation of the noise for each frame. Quantitative analysis was performed using volume weighted radial plots on the remaining 655 patients and normals using both prefilters. Corresponding accuracy for each technique was determined by calculating the area under the receiver operating curve. Results: There was a small but statistically significant increase in accuracy using the Coiflet prefilter. The area under the ROC curve was 0.81±0.02 using the conventional Butterworth prefilter versus 0.84±0.02 using the Coiflet prefilter (p = 0.02). Conclusions: Wavelet prefiltering results in a small but statistically significant increase in accuracy of quantitative SPECT thallium-201 myocardial perfusion imaging.