Abstract
261
Objectives Standard MRI segmentation-based attenuation correction (MRAC) of PET data ignores the inter/intra-patient variability of linear attenuation coefficients (LACs) in each tissue class, which can lead to quantification errors, especially in the lungs. In this work, we derive continuous and patient-specific LACs from time-of-flight (TOF) emission data using a constrained maximum likelihood reconstruction of attenuation and activity (MLAA) algorithm.
Methods Standard 4-class MRAC maps were derived from the segmentation of CT images of 19 18F-FDG (n = 17) and 18F-Choline (n = 2) PET/CT patients acquired on the Siemens mCT scanner into background air, lung, fat and soft tissue classes, followed by the assignment of predefined LACs of 0, 0.0224, 0.0864 and 0.0975 cm-1, respectively, to each class. Lung LACs of the MRAC maps were then estimated from emission data using an MLAA algorithm constrained by Gaussian lung tissue preference and Markov random field smoothness priors. Quantification performance of the MRAC and MLAA-AC methods was compared to the reference CTAC method in the lungs, lesions in/near the lungs and adjacent organs.
Results The results showed that the proposed MLAA method can retrieve lung density gradients and fairly compensate for respiratory-phase mismatch between PET and attenuation maps. The mean of the estimated lung LACs generally follow the trend of the reference CTAC method. Quantitative analysis revealed that the MRAC method results in average errors of -5.2 ± 7.1% and -6.1 ± 6.7%, whereas the MLAA algorithm reduced the errors to -0.8 ± 6.3% and -3.3 ± 4.7% in lungs and lesions, respectively.
Conclusions The proposed constrained MLAA algorithm is promising in deriving patient-specific lung LACs for attenuation correction in TOF PET/MR imaging. The clinical adoption of such an approach is feasible and straightforward.
Research Support This work was supported in part by the Swiss National Science Foundation under Grant SNSF 31003A-149957 and by the Indo-Swiss Joint Research Programme ISJRP 138866.