A sparse transmission method for PET attenuation correction in the head

We describe a novel solution for PET attenuation correction in the head based on the joint reconstruction of simultaneously acquired emission and sparse transmission (sTX) data. We demonstrate that an sTX array can provide better cross-talk reduction than a conventional full-ring transmission source. This initial evaluation is based on synthetic 2D non-time-of-flight data corresponding to 20 fixed line sources placed in a 30 cm diameter ring around the head. A 57.2 cm ring diameter is also considered. The total source activity equals half the total emission activity in the brain. Simultaneous emission/transmission and blank scans are simulated. These data are iteratively reconstructed to estimate both the emission and linear attenuation coefficient (LAC) images of the head. We find that the sTX data effectively constrain cross-talk. Bone, soft tissue and voids are approximately represented in the estimated attenuation image. The results are compared to a continuous ring-source-based joint reconstruction, as well as to a standard MLEM reconstruction of emission-only data assuming a uniform LAC distribution within the head. 10-20% underestimation of activity in the peripheral regions of the brain in the latter two images is reduced to <; 5%on average in the sTX case. Initial tests indicate the algorithm is robust to a realistic level of noise. We estimate that such an sTX technique would result in a negligible increase in patient absorbed radiation dose in a typical 18FDG clinical study of the head.