Abstract
648
Objectives We compare the quantitative accuracy of MRI-based attenuation correction using the 3-class attenuation map (PET-MRAC3c) implemented on the Ingenuity TF PET/MRI and the 4-class attenuation map (PET-MRAC4c) implemented on the Siemens mMR using clinical studies. CT-based attenuation corrected PET images (PET-CTAC) were used as reference.
Methods Fourteen patients underwent whole body sequential 18F-FDG PET/CT and PET/MR imaging. T1-weighted MR images were segmented according to the technique used on the Ingenuity TFI to obtain a 3-class attenuation map followed by assignment of theoretical attenuation coefficients (air: 0 cm-1, lung: 0.022 cm-1, soft-tissue: 0.098 cm-1). A 4-class attenuation map was also derived from the Dixon sequence using a technique similar to the one used on the Siemens mMR (air: 0 cm-1, lung: 0.018 cm-1, fat: 0.086 cm-1, soft-tissue: 0.098 cm-1). SUVmean and SUVmax metrics were calculated for twelve volumes of interest defined in various organs/tissues and malignant lesions.
Results PET-MRAC3c and PET-MRAC4c revealed significant underestimation of SUV for normal organs (-17.4±8.5% and -22.0±6.8%, respectively) compared to PET-CTAC used as reference. Lesion’ SUVs presented the same trend with larger underestimation for PET-MRAC4c (-9.2±6.1%) compared to PET-MRAC3c (-3.9±9.0). The different attenuation coefficients assigned to the lungs between both techniques resulted in significant positive bias in PET-MRAC3c (18.6±15.3%) and fairly low negative bias in PET-MRAC4c (-0.5±13.3%) in this region. The largest differences were observed in bony structures or near massive bone tissue for both approaches. Despite the large bias, there was good correlation between PET-MRAC3c (R=0.97, p<0.01) and PET-CTAC and PET-MRAC4c (R=0.97, p<0.01) and PET-CTAC.
Conclusions PET-MRAC3c resulted in significant systematic positive bias in the lung region, whereas a negative bias was observed in other regions. PET-MRAC4c slightly underestimated tracer uptake in the lungs. Nevertheless, it resulted in considerable negative bias, even larger than that of PET-MRAC3c, in other regions.