Embedding Partial Volume Correction in Atlas-Based Image Analysis for Substriatal Uptake in Dopaminergic PET Imaging

3001 Objectives: PET is an invaluable molecular imaging tool that visualizes the dopaminergic system. Quantitative measurements of nigrostriatal dopamine functions are beneficial for the differential diagnosis of Parkinsonism. Subregional analysis of striatal uptake yields a powerful diagnostic performance. However, partial volume effect (PVE) hinders quantitative accuracy, which is significant for small structures. This work proposes a partial volume correction (PVC) strategy that is embedded in atlas-based image analysis to quantify substriatal uptake in dopaminergic PET imaging. Methods: Monte Carlo simulations were performed on an MRI-based brain phantom that models realistically healthy and neurodegenerative situations to assess the performance of the proposed method. Simulated PET images were co-registered to the corresponding MR images. The MR images were spatially normalized to the atlas template supplied by the Montreal Neurological Institute (MNI; Montreal, Canada) within the statistical parametric mapping package (Wellcome Department of Cognitive Neurology, London, UK). Subsequently, MR-matched PET images were converted into the MNI space using the previous transformation parameters. Normalized PET images were corrected for the PVE according to the recovery coefficients that were derived from a combination of traditional and large regions of interest (ROIs). A set of atlas-derived ROIs was carefully defined only once and applied to quantify the specific binding (SB) for the whole striatum and its subdivisions. Results: For the healthy state, SB values before PVC were underestimated to be 33.8% for the caudate nucleus (CA), 23.0% for the anterior putamen (APU), 34.0% for the posterior putamen (PPU), 28.5% for the entire putamen (PU), and 30.3% for the striatum (ST). After PVC, the biases were 0.3%, 0.3%, 0.8%, 0.3%, and 0.8% for SBCA, SBAPU, SBPPU, SBPU, and SBST, respectively. For the simulation of neurodegeneration, the SB values before PVC were underestimated to be 31.4% for CA, 24.3% for APU, 24.8% for PPU, 25.5% for PU, and 28.1% for ST. The proposed PVC approach yielded respective biases of 1.1%, 6.3%, 8.4%, 0.4%, and 2.0% for SBCA, SBAPU, SBPPU, SBPU, and SBST. The respective average percentage errors before PVC were 29.9% and 26.8% for healthy and neurodegenerative states; the errors after PVC for healthy and neurodegeneration were 0.5% and 3.6%, respectively. Conclusions: We implemented an approach that automatically analyzes images and corrects PVE to accurately quantify the substriatal uptake for dopaminergic PET images. Ongoing works include evaluation of the proposed technique in an actual patient population.