Search

Background: The correction of γ-photon attenuation in PET-MRI remains a critical issue, especially for bone attenuation. This problem is of great importance for brain studies due to the density of the skull. Current techniques for skull attenuation correction (AC) provide indirect estimates of cortical bone density, leading to inaccurate estimates of brain activity. The purpose of this study was to develop an alternate method for bone attenuation correction based on NMR. The proposed approach relies on the detection of hydroxyapatite crystals by zero echo time (ZTE) MRI of P, providing individual and quantitative assessment of bone density. This work presents a proof of concept of this approach. The first step of the method is a calibration experiment to determine the conversion relationship between the P signal and the linear attenuation coefficient μ. Then P-ZTE was performed in vivo in rodent to estimate the μ-map of the skull. F-FDG PET data were acquired in the same animal and reconstructed with three different AC methods: P-based AC, AC neglecting the bone and the gold standard, CT-based AC, used to comparison for the other two methods. Results: The calibration experiment provided a conversion factor of P signal into μ. In vivo P-ZTE made it possible to acquire 3D images of the rat skull. Brain PET images showed underestimation of F activity in peripheral regions close to the skull when AC neglected the bone (as compared with CT-based AC). The use of P-derived μ-map for AC leads to increased peripheral activity, and therefore a global overestimation of brain F activity. Conclusions: In vivo P-ZTE MRI of hydroxyapatite provides μ-map of the skull, which can be used for attenuation correction of F-FDG PET images. This study is limited by several intrinsic biases associated with the size of the rat brain, which are unlikely to affect human data on a clinical PET-MRI system. [ABSTRACT FROM AUTHOR]

Most current brain metabolic models are not capable of taking into account the dynamic isotopomer information available from fine structure multiplets in C spectra, due to the difficulty of implementing such models. Here we present a new approach that allows automatic implementation of multi-compartment metabolic models capable of fitting any number of C isotopomer curves in the brain. The new automated approach also makes it possible to quickly modify and test new models to best describe the experimental data. We demonstrate the power of the new approach by testing the effect of adding separate pyruvate pools in astrocytes and neurons, and adding a vesicular neuronal glutamate pool. Including both changes reduced the global fit residual by half and pointed to dilution of label prior to entry into the astrocytic TCA cycle as the main source of glutamine dilution. The glutamate-glutamine cycle rate was particularly sensitive to changes in the model. [ABSTRACT FROM AUTHOR]