Abstract
This paper presents a fully automatic computer aided diagnosis (CAD) system for the classification of Parkinson's disease (PD) by means of functional imaging, such as, the single photon emission computed tomography (SPECT). Firstly, in the preprocessing step, Histogram Equalization (HE) is applied on all the 3D SPECT image data. Secondly, HE is applied on the so-called non-specific (NS) region, as reference region. Then, the normalized images are modelled using Principal Component Analysis (PCA). Thus, for each subject, its scan is represented by a few components. These resulting features will be used for the classification task. The proposed system has been tested on a 269 image database from the Parkinson Progression Markers Initiative (PPMI). Classification rate of 92.63% is achieved, which has proved the robustness and the productiveness of the proposed CAD system in PD pattern detection. In addition, the PCA based feature extraction approach significantly improves the baseline Voxels-as-Features (VAF) method, used as an approximation of the visual analysis. Finally, the proposed aided diagnosis system outperforms several other recently developed PD CAD systems.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call