Quantitative Analysis of DAT SPECT Images by 3D Striatum Recognition

Objective: Brain dopaminergic neurotransmission is currently evaluated by SPECT with dopamine transporter (DAT) ligands, with 123I FP‐CIT (DaTSCAN ‐ GE Healthcare) being the most utilized. Visual assessment of the DAT SPECT images, which depict the quantity of radioactivity absorbed by the striatum, is often used and considered sufficient in many diagnostic situations of Parkinson Disease. Most of all the existing processing methods of DAT SPECT perform qualitative or semi‐ quantitative analysis and allow medical doctors to easily distinguish Parkinson’s syndrome from essential tremor, but do not permit an exact assessment of the disease progression and the drugs’ effects because of the high intra‐observer and inter‐observer variability due to the manual intervention both for the selection of the slices where the absorbed radioactivity is more visible and for the positioning of the ROIs (region of interest) in the selected slices. Therefore, the objective of this work is to propose a system that performs a quantitative analysis making the DAT‐SPECT examination fully reproducible. Method: We propose an effective 3D Striatum reconstruction method combined with pre‐ and post‐filtering for the automatic striatum extraction from SPECT images for performing quantitative and reliable measurements. In detail, the algorithm initially performs a pre‐filtering phase in order to remove noise due to the acquisition step and then it automatically extracts all the slices where the striatum is present and finally reconstructs the striatum volume. The 3D reconstruction algorithm builds a 3D model representing the surface of interest, starting from a set of SPECT slices. The algorithm builds an approximated 3D model and iteratively modifies it by a optimization method. During each iteration, the position of every single vertex of the model is changed according to a composition of external and internal forces of the images. Finally, in order to increase the system’s accuracy a correction module for partial volume effects (PVEs) is developed. Conclusions: The proposed algorithm automatically performs the 3D segmentation of striatum. We have already collected a set of 200 patients together with the clinical diagnosis. We plan to estimate the error in the final assessment due to qualitative methods over the considered set of patients and to analyze the reproducibility of the automated method.