A MRI-based PET attenuation correction with μ-values measured by a fixed-position radiation source

Several MRI-based attenuation correction methods have been reported for PET/MRI. The accuracy of the attenuation map (μ-map) from an MRI image depends on correctness of the segmentation of tissue and the attenuation coefficients to be assigned (μ-values). However, an MRI image does not reflect the attenuation of radiation and inaccurate assignment of μ-values affects the quantitative assessment of functional images of PET. Although installation of a transmission scan function on the PET/MRI can provide an accurate μ-map, it restricts the design of the scanner, increases the manufacturing cost and takes additional scanning time. In this study, we implemented an MRI-based μ-value estimation method with a non-rotational radiation source to construct the proper μ-map for PET/MRI and assessed it based on clinical data sets. The proposed method uses the accurately segmented tissue map, the partial path length of each tissue, and detected intensities of attenuated radiation from a fixed-position radiation source which usually rotates around the subject to obtain the μ-map with the tomographic procedure. According to the Lambert-Beer law, attenuated intensity is described as the function of partial path length and μ-values of every tissue. The partial path length could be estimated by the simulation of fixed-point radiation with the same scanner geometry using the known tissue map from MRI. The μ-values of every tissue could be estimated by inverting the function. The simulation results, based upon measurement data, showed the errors between μ-values of the conventional transmission scan and our proposed method were 2.3±0.9%, 18.6±8.0% and -11.1±5.5% for brain, bone and soft tissue other than brain, respectively. Although there were over- and under-estimations for bone and soft tissue, respectively, the present method is able to estimate the brain μ-value accurately in clinical situations and that strongly affects the quantitative value of PET images because of the large volumetric ratio of the brain.