Reconstruction of uniform sensitivity emission image with partially known axial attenuation information in PET-CT scanners

In PET-CT the axial length of image reconstruction is defined by the CT scan, which delivers an axial extend-dependent radiation dose. The beginning and end scanning points for CT and therefore PET scans are typically chosen in such a way that the PET scan is performed with a particular number of beds. While PET bed overlapping is optimized to achieve uniform image sensitivity, the whole volume edge planes suffer from low sensitivity, since only direct plane LORs are available from the acquisition. This problem can be solved through object over scanning by additional bed acquisitions. This does not result in an additional PET radiation dose. Still, the edge planes will have lower sensitivity due to the absence of oblique LOR attenuation factors. ACFs are necessary for sensitivity restoration of oblique LORs and are accumulated over the attenuation map image beyond CT scanning points. Recent theoretical investigations concluded that both activity and attenuation distributions can be obtained from PET emission TOF data alone with knowledge of the sinogram scaling parameter. In this work we consider an easier problem for the above-mentioned application. ACFs, converted to 511 energy units, are partially known and scatter distribution is already estimated based on direct plane LORs. The iterative algorithm to reconstruct only the emission image, with estimation of partially unknown ACFs, is presented. The estimation of ACFs is a sub-product of emission activity reconstruction. The algorithm is of a similar complexity to commonly used ML-EM. Initial investigations of experimental data show that partial ACF knowledge restores attenuation information necessary for the uniform sensitivity of PET reconstructed volume edge planes.