Attenuation correction of cardiac perfusion images in SPECT using Compton scatter data: a Monte Carlo investigation

Attenuation correction of SPECT cardiac perfusion images requires a patient-specific attenuation map. Such a map can be estimated from segmenting both the Compton scatter and photopeak data to identify the regions of the lungs and other soft tissues, and assigning appropriate attenuation coefficients to the regions. Using simulated images produced by Monte Carlo techniques, we have previously demonstrated that a reasonable segmentation can be obtained. This work was extended to investigate the relative accuracy of attenuation correction applied to Monte Carlo images using such maps, with respect to correction with the true attenuation maps. Results demonstrated that the regions of the lungs segmented in the Compton scatter window images were smaller in size than the true lungs by about 17 to 28%. However, this error did not create a significant bias in the polar maps. The bias was seen primarily in the anterior septal region near the base of the heart