Visual Detection of Parkinson's Disease via Facial Features Recognition
【摘要】：Parkinson's disease(PD) is a widely concerned neurodegenerative disease, due to its high prevalence in the elderly population. In this paper, a simple automated framework for PD detection is designed. The framework is proposed by extracting geometric and texture features based on facial visual information. Geometric features are defined via interrelationships of the facial landmarks, indicating the physical information of facial features. Texture features are extracted with a method fused with Local Binary Pattern(LBP) and Gray-level Co-occurrence Matrix(GLCM) that enhance the discriminative ability of the system, indicating the digital information of facial features. The proposed method performs well with a variety of classifiers, and the best result reaches 0.936 accuracy and 0.938 F_1 score. Results indicate this simple automated framework for PD detection could provide a valuable tool for the clinical assessment of PD, which could assist in PD screening effectively.