Is transmission-gating necessary for cardiac SPECT imaging with attenuation correction?

摘要

Accurate attenuation correction (AC) requires matched emission (EM) and transmission (TX) data. For cardiac SPECT imaging with AC, historically TX data is only acquired in a non-gated manner and AC is only performed on the summed EM data. Considering that the reconstruction of the summed EM data can be decomposed into the reconstruction of each gated bin, AC using the same non-gated TX data leads to a mismatch of the EM and TX data because the volume, shape, and/or location of the heart can vary for each gated bin in the EM data, yet the same non-gated TX data is used for AC. Analytically, the FBP image (without AC) is the superposition of the true image scaled by a scale image and a shift image. The scale image is solely determined by and can be computed from the attenuation map and the emission acquisition geometry. The shift image is due to the lack of AC and becomes zero everywhere if AC is performed accurately. Using the XCAT phantom, we generated non-gated, heart beat gated and respiratory motion gated attenuation maps. We then computed and compared the scale images for the left ventricle wall at each gated bin using the non-gated and gated attenuation maps. Using a patient study that showed smaller cardiac size at stress than at rest in both EM and respiration-averaged low-dose CT images, we reconstructed the stress image with AC using the stress CT image then the rest CT image, mimicking AC of a gated bin with matched gated TX data and non-gated TX data. Results show that AC using gated versus non-gated attenuation maps differed by a magnitude of 6% from bin to bin for heart beat gating and 12% for respiratory motion gating, but the difference averaged over all the bins was within 1%. For comparison, AC using an attenuation map that was 6.5 mm misregistered with emission image (heart in emission was positioned 6.5 mm into the left lung) showed a 10% difference as compared to without misregistration. The stress images of the patient with AC using the stress a--nd rest CT images showed maximal segment perfusion difference of 6%. The results suggested transmission gating was necessary for accurate AC of each gated bin. However, if the emission images of the gated bins were aligned properly and summed to produce the final perfusion image, the difference was averaged out and became minimal. Transmission gating was hence not necessary.