Basal Ganglia Activity Measurement by Automatic 3-D Striatum Segmentation in SPECT Images

In this paper, we propose an automatic system for quantitative measurements of basal ganglia activity by 3-D striatum activity reconstruction from 123 I FP-CIT SPECT images for investigations on the nigrostriatal dopaminergic system. In detail, the proposed system consists of three main parts: 1) Image Enhancement, carried out by means of Active Contours Models, 2) 2-D Striatum segmentation in the HSV color space, and 3) Surface Rendering by an extended marching cubes algorithm where the isovalue of the striatum's isosurface and the final 3-D volume are automatically computed. The validation of the model was done on the anthropomorphic Alderson RSD phantom, achieving encouraging results both in terms of striatum activity measurement (average error of 1.57% ± 1.28%) and in terms of volume measurements (average error of 0.77% ± 1.42%). Finally, an evaluation on 128 patients was carried out, suggesting novel ranges (to be further clinically validated) of the reference values for striatum activity that may be employed for the differential diagnosis of Parkinson, Vascular Parkinsonism, Alzheimer, and Essential Tremor.