Direct Evaluation of MR-Derived Attenuation Correction Maps for PET/MR of the Mouse Myocardium

Attenuation correction (AC) must be applied to provide accurate measurements of PET tracer activity concentrations. Due to the limited space available in PET/MR scanners, MR-derived AC (MRAC) is used as a substitute for transmission source scanning. In preclinical PET/MR, there has been limited exploration of MRAC, as the magnitude of AC in murine imaging is much smaller than that required in clinical scans. We investigated if a simple 2 class (air and tissue) segmentation-based MRAC approach could provide adequate AC for mouse PET imaging. To construct the default MRAC ${\mu }$ maps, MR images were thresholded and segmented using ASIPRO software (Siemens Molecular Imaging), which defined the mouse body region as tissue with a uniform linear attenuation coefficient ( ${\mu }$ ) of $0.095~\hbox{cm}^{ - 1}$ , and the background and lungs as air, with a ${\mu }$ value of $0~\hbox{cm}^{ - 1}$ . To correct for the misassignment of the lungs as air, two further MRAC ${\mu }$ maps were tested: 1) MRAC (tissue) approach, which changed the lung region designation from air to tissue ( ${\mu } = 0.095~\hbox{cm}^{ - 1}$ ) and 2) MRAC (lung) approach, which treated the lungs as an additional tissue class, with a ${\mu }$ value of $0.032~\hbox{cm}^{ - 1}$ . All ${\mu }$ maps were then forward projected to create attenuation sinograms for image reconstruction. Standard uptake value (SUV) maps of the myocardium were derived for 10 mice with and without AC applied using gold standard transmission scans (TXAC), the 3 MRAC methods and PET emission scans (EmAC). All AC methods produced significantly different myocardial SUVs to those produced without AC when compared across the mouse group ( ${\rm p} ). Similar ( $ ) SUV were derived with all AC methods, with the best agreement to TXAC achieved using the MRAC (tissue) method, giving a mean difference of $0.9{\pm }2.4\% $ in myocardial SUV when compared across all mice. SUV differences of up to 40%, however, were seen in areas adjacent to the RF coil in images produced using all AC methods, except for TXAC. A 2 class MRAC approach can therefore provide acceptable AC for myocardial imaging in mice, although additional CT templates of coils and animals beds would be recommended to further improve image quantification.