Comparison Between Affine and Non-affine Transformations Applied to I~(123)-FP-CIT SPECT Images Used for Parkinson's Disease Diagnosis

摘要

In recent years, the use of I~([123])-FP-CIT or I~([123])-Ioflupane SPECT images has emerged as an effective support tool for Parkinson's Disease diagnosis. Many works in this field have consisted on comparing different images obtained from subjects both Healthy Control (HC) subjects and patients with Parkinsonism (PD) and using them to obtain measures (features) able to discern among them. In this scenario, spatial normalization of I~([123])-FP-CIT images is fundamental to match equivalent areas of the brain from different subjects. This work tries to compare the two most common ways to make the spatial normalization of SPECT images from PD and HC subjects in the study of Parkinsonism: affine and non-affine transformations. For that, these two approaches have been applied to a set of 20 images obtained from 20 different subjects (11 HC and 9 with PD) and measured how volume of new voxels, when applying normalization to a reference template, has changed. Despite the accurate match obtained when using a non-affine spatial normalization procedure, using this method involves that some parts of the brain are compressed or stretched in excess to fit the template. This effect is even more pronounced when using PD images than HC. Using the affine procedure, striatum area preserves better its morphology and can be used to obtain more reliable morphological features.