Neurodegenerative diseases detection using distance metrics and sparse coding: A new perspective on gait symmetric features

• Published in: Computers in biology and medicine Vol. 120; p. 103736
• Main Authors: Ghaderyan, Peyvand, Ghoreshi Beyrami, Seyede Marziyeh
• Format: Journal Article
• Language: English
• Published: United States Elsevier Ltd 01.05.2020; Elsevier Limited
• Subjects: Algorithms; Amyotrophic lateral sclerosis; Chebyshev approximation; Classification; Classifiers; Coding; Datasets; Divergence; Dynamic gait series warping; Euclidean geometry; Gait; Huntington's disease; Huntingtons disease; Internal Medicine; Methods; Movement disorders; Neural coding; Neurodegenerative diseases; Other; Parkinson's disease; Principal components analysis; Signal processing; Symmetry; Time series; Trigonometric functions; Amyotrophic lateral sclerosis; Parkinson's disease; Dynamic gait series warping; Huntington's disease
• ISSN: 0010-4825; 1879-0534; 1879-0534
• DOI: 10.1016/j.compbiomed.2020.103736

Gait rhythm fluctuations are of great importance for automatic neurodegenerative diseases (NDDs) detection. They provide a cost-effective and noninvasive monitoring tool in which their parameters are related to neuromuscular function. This study investigated a new solution based on a set of new symmetric features and sparse non-negative least squares (NNLS) coding classifier. Dynamic gait series warping (DGSW), Euclidean, Manhattan, Minkowski, Chebyshev, Canberra distances, and cosine function were used to quantify the amount of divergence between the left and right stride, swing, and stance intervals. The algorithm was evaluated using the gait signals of 16 healthy control subjects, 13 patients with amyotrophic lateral sclerosis (ALS), 15 patients with Parkinson's disease (PD) and 20 patients with Huntington's disease (HD). The proposed new approach using symmetric features and NNLS technique achieved outstanding accuracies of 98%, 97%, and 95% on the patients with PD, ALS, and HD, respectively. The findings also suggested that the new DGSW, cosine function, and Chebyshev distance, which are designed to dynamically, geometrically, or nonlinearly quantify the similarity between two time series, provide the discriminatory measures to describe how NDDs alter the gait symmetry. In comparison with other studies, combining symmetric features with a sparse NNLS coding classifier can improve the detection accuracy providing an efficient and cost-effective framework for the development of a NDDs detection system. •A set of new distance-based gait symmetric feature is developed for NDDs detection.•Dynamic gait series warping is used to nonlinearly quantify inter-limb deviations.•Compared to SVM, the sparse coding method shows improved performance.