Tissue Probability-Based Attenuation Correction for Brain PET/MR by Using SPM8

Bone attenuation remains a methodological challenge in hybrid PET/MR, as bone is hard to visualize via magnetic resonance imaging (MRI). Therefore, novel methods for taking into account bone attenuation in MR-based attenuation correction (MRAC) are needed. In this study, we propose a tissue-probability based attenuation correction (TPB-AC), which employs the commonly available neurological toolbox SPM8, to derive a subject-specific $\mu$ -map by segmentation of T1-weighted MR images. The procedures to derive a $\mu$ -map representing soft tissue, air and bone from the New Segment function in SPM8 and MATLAB are described. Visual and quantitative comparisons against CT-based attenuation correction (CTAC) data were performed using two $\mu$ -values ( $0.135~\hbox{cm}^{ - 1}$ and $0.145~\hbox{cm}^{ - 1}$ ) for bone. Results show improvement of visual quality and quantitative accuracy of positron emission tomography (PET) images when TPB-AC $\mu$ -map is used in PET/MR image reconstruction. Underestimation in PET images was decreased by an average of $5 \pm 2$ percent in the whole brain across all patients. In addition, the method performed well when compared to CTAC, with maximum differences (mean $ \pm$ standard deviation) of $ - 3 \pm 2$ percent and $2 \pm 4$ percent in two regions out of 28. Finally, the method is simple and computationally efficient, offering a promising platform for further development. Therefore, a subject-specific MR-based $\mu$ -map can be derived only from the tissue probability maps from the New Segment function of SPM8.