A computer aided diagnosis system for the early detection of neurodegenerative diseases using linear and non-linear analysis

• Published in: 2018 IEEE 4th Middle East Conference on Biomedical Engineering (MECBME) pp. 116 - 121
• Main Authors: Elden, Rana Hossam, Ghoneim, Vidan Fathi, Al-Atabany, Walid
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.03.2018
• Subjects: Amyotrophic lateral sclerosis; Biomedical engineering; Diseases; Entropy; Feature extraction; Fisher score selection strategy; Foot; Gait time series; Huntington; Linear features; Non-linear features; Parkinson; Support vector machine; Support vector machines; Time series analysis
• ISSN: 2165-4255
• DOI: 10.1109/MECBME.2018.8402417

One of the most serious problems that faces human nowadays is the gait disturbances as result of neurodegenerative diseases (NDD). Neurodegenerative diseases such as Parkinson's disease (PD), Amyotrophic Lateral Sclerosis (ALS), Huntington Disease (HD) identified as the dynamic loss of neurons in human brain. Therefore, gait analysis can yield a significant approach for the early diagnosis of gait disturbances and determine the treatment plan with the generation of new era of computerized medical systems for analyzing such diseases. The present study explores the improvement of the classification capability by using non-linear features with previously used linear features. Fisher score selection strategy is used to get the optimal feature subset and the optimal gait time series in classifying NDD. Support vector machine (SVM) with radial basis kernel function (RBF) is implemented for discriminating NDD patients against healthy ones optimized by leave-one-out-cross-validation (LOOCV). The applied classifier differentiated NDD subjects from healthy ones with an area under the receiver operating characteristic curve "0.861" and an overall accuracy "90.625%".