Practical Use of Alpha-Shapes in Neuroimaging for Parkinson's Disease Diagnosis

Parkinson's disease (PD) is a progressive neurological disorder characterized by tremor, ridigity, bradykinesia and cognitive impairment. In clinical practice, the FP-CIT SPECT scans are one of the most widely used tools to diagnose PD and also to measure its progression. Whereas Healthy Control (HC) subjects show their striatum region highly illuminated and with a c-shape, in patients with PD this area is not as much highlighted and it presents a rounded shape. Traditionally, analysis of FP-CIT scans have been carried out following voxel-wise comparisons between HC and PD subjects. This approach is limited due to the necessity of non-linear spatial registrations that might alter artificially the interclass separation between HC and PD subjects due to the intrinsic deformation of striatum region in PD scans. In this scenario, morphological analyses have become a more effective alternative for determining this separation. In this work we analyze the contour of 208 FP-CIT SPECT volumes to obtain a set of features that will support PD diagnosis. For this purpose, 3D alpha-shapes have been used to measure this contour. In addition, we propose a method to determine the intensity threshold from which most of the information is contained. This helps to limit the number of input features and eliminates potential background noise.