Abstract
1981
Objectives The accuracy of nuclear stress testing with single photon emission computed tomography (SPECT) myocardial perfusion imaging is diminished in the presence of attenuation artifacts. Repeat imaging of the patient in prone rather than supine position, or reprocessing the data using computed tomography-based attenuation correction (AC) algorithms, often lessens these artifacts, but may create new ones as well. This leaves the onus on the reader to mentally integrate various data sets. To facilitate more accurate and rapid interpretation, a novel imaging processing method that generates a composite image of supine, prone, and AC datasets was developed and evaluated.
Methods In this IRB-approved study, reconstructed stress myocardial perfusion SPECT images were obtained from patients for both supine and prone patient orientations without AC and supine with AC. Potentially abnormal studies with possible perfusion abnormalities on supine non-AC images underwent reprocessing to generate a composite image. The first step was image registration of the three datasets using a 3D rigid body transformation based on the computed center-of-mass of the myocardial image intensities. The next step was to normalize the image intensities across datasets. Two normalization methods were considered: one based on the slice-by-slice maximum intensity in the myocardium, the other based on the global maximum. Finally, average intensity and maximum intensity composite images were created by assigning an intensity to each pixel in the composite image that represented either the average or maximal intensity of that pixel across the three datasets.
Results The method was evaluated using a preliminary set of five studies. Inferior defects attributable to diaphragmatic attenuation on supine non-AC images were generally resolved by prone non-AC imaging and supine AC imaging, although composite images were more consistent. Apical perfusion defects likely due to overcorrection on AC images were resolved by the composite images. Readers preferred the composite image format for interpretation. The processing method most effective was the slice-by-slice normalization method along with the maximal intensity composite image.
Conclusions An image processing method that generates a composite image from the supine no AC, prone no AC and supine AC imaging protocols demonstrated the ability to resolve conflicting results from the individual protocols and was preferred by readers.