Evaluation of an attenuation correction method for thallium-201 myocardial perfusion tomographic imaging of patients with low likelihood of coronary artery disease

BACKGROUND

Image artifacts caused by nonuniform photon attenuation are a source of error in interpretation of images during myocardial perfusion single photon emission computed tomography (SPECT). A newly introduced attenuation correction method was evaluated for improvement in image homogeneity during 201Tl SPECT. The method was assessed with a cardiac phantom and in examinations of 42 patients (29 men) with a low likelihood of coronary disease.

METHODS AND RESULTS

Simultaneous transmission-emission SPECT was performed with a moving collimated 153Gd line source synchronized with a moving electronic acquisition window for transmission imaging and a novel variable-width electronic exclusion window for emission imaging designed to avoid transmission-to-emission cross talk. The resulting uncorrected and corrected polar maps were analyzed visually and divided into 31 segments for quantitative analysis. Visual analysis of the color-coded mean polar maps showed clear improvement in homogeneity after correction among the phantom, male patients, female patients, and 42 patients combined at stress and redistribution. The male and female mean polar maps showed very little differences in regional count distribution after correction. Quantitative analysis of the mean polar maps showed the following mean segmental counts (%SD) before and after attenuation correction: phantom 88 (9) to 90 (7.5), P = .00005; men at stress 83 (10) to 88 (6), P = .0007, and at redistribution 84 (8) to 88 (6), P = .01; women at stress 86 (7) to 90 (5), P = .0002, and at redistribution 87 (5) to 88 (7), P = .3; patients combined at stress 84 (8) to 88 (6), P = .0004, and at redistribution 85 (7) to 87 (7), P = .03. Inferior/anterior count ratio for men at stress increased after correction from 0.82 to 0.99 and septal/lateral count ratio from 0.94 to 1.02. Inferior/anterior count ratio for men at redistribution increased from 0.86 to 1.06 and septal/lateral count ratio from 0.97 to 1.04. Inferior/anterior count ratio for women at stress increased from 0.95 to 1.03 and septal/lateral count ratio from 0.93 to 1.00. Inferior/anterior count ratio for women at redistribution increased from 1.04 to 1.10, and septal/lateral count ratio decreased from 1.02 to 1.00.

CONCLUSION

Improvement in image homogeneity was demonstrated with this attenuation correction method with a cardiac phantom and for patients with low likelihood of coronary artery disease. The slight relative increase in inferior wall counts at redistribution was most likely caused by scatter from the relatively higher liver activity compared with the situation during stress and emphasizes the need for scatter correction. The close similarity in count distribution for the mean male and female polar maps supports use of a sex-independent normal database for quantitative analysis. The reduced variation in corrected images from patient to patient implies increased accuracy for detection of myocardial defects.

JET CLASSIFICATION AT CDF

Results for the classification of jets obtained with the use feed-forward neural networks are reviewed with particular attention to model-dependence, flavor sensitivity and Et-dependence. First results obtained with a Kohonen-type network are also presented and both are compared with those obtained with a Fisher discriminant.