Early Diagnosis of Parkinson’s Disease using Orientation Texture Features

Abstract

This study focuses on the identification of Parkinson’s Disease (PD) at an early stage using
intensity normalization technique. The normalization techniques plays a vital role in the medical image
analysis. Identification of an appropriate technique is paramount in the initial stage in Computer Aided
Diagnosis (CAD) system. The alpha stable distribution technique is applied for suitable normalization of
three-dimensional SPECT (Single Photon Emission Computed Tomography) images. The orientation
texture features of GLCM (Gray-Level Co-Occurrence Matrix) feature set are extracted from SPECT
images along with surface fitting based features, shape features for auto classification of ELM (Extreme
Learning machine) classifier. All these feature models are provided and their optimal sets are evaluated
based on the consistency of the attributes. The application of alpha stable distribution based intensity
normalization technique and orientation texture features to the experimental test of the proposed system has
brought about a remarkable performance. The use of performance and consistency measures has ensured the
validity of the experiments. The kernel based ELM classifier confirms the best performance accuracy of
90.45%. The proposed method is advantageous to neurologist in the early diagnostic process of Parkinson’s
disease