Sensitivity and Attenuation Weighted Multi-Bed PET Acquisition and Reconstruction

The total number of counts detected in a PET scan is a function of the 3D sensitivity (S) of the scanner, the scan time (T) combined with the attenuation (A) of the emitted photons along the line of response (LOR). Varying the scan time per bed could provide uniform noise properties along the axial direction in a multibed study. In this work, we use the sensitivity weights of the PET scanner and attenuation (SA) to determine the scan time required per bed for a multi-bed step-and-shoot study so as to generate a 3D reconstructed image with uniform noise properties. Though we talk about step-and-shoot studies in this work the same concept can be easily expanded for PET scans acquired using Continuous Bed Motion (CBM). Assuming the total scan time for the multi-bed study to be constant, the variable scan time per bed using SA can be quickly calculated as soon as the attenuation map of the patient is available. For e.g. the cranial region and legs could have less attenuation compared to the torso of a large patient. Multibed studies were simulated to scan from head to thorax region using an eight ring Vision PET (Siemens Medical Solutions USA, Inc.) geometry. Two scan modes were evaluated: (1) Same scan time in all 3 beds and (2) Scan such that the SAT is constant at the center of each bed. To determine the SAT, the attenuation weighted expanded norm sinogram was back projected to the image space and the voxels within the patient boundary were used to calculate the slice by slice SA along the axial direction. Based on the SA the relative scan time per bed was determined. Reconstructed images obtained using uniform SAT was found to provide more uniform noise properties compared to those obtained using same scan time per bed. Finally, the SAT weights were also used to stitch the images together for multibed studies.