Classification of neurodegenerative diseases using gait dynamics via deterministic learning

• Published in: Information sciences Vol. 317; pp. 246 - 258
• Main Authors: Zeng, Wei, Wang, Cong
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 01.10.2015
• Subjects: Classification; Constants; Deterministic learning; Dynamics; Errors; Estimators; Gait; Gait analysis; Gait dynamics; Movement disorders; Networks; Neurodegenerative diseases; Training; Gait dynamics; Neurodegenerative diseases; Deterministic learning; Gait analysis; Movement disorders
• ISSN: 0020-0255; 1872-6291
• DOI: 10.1016/j.ins.2015.04.047

•We present a new method to classify neurodegenerative diseases via deterministic learning theory.•The gait system dynamics can be learned by using RBF neural networks.•The neurodegenerative diseases can be classified according to the smallest error principle.•The discriminability provided by the dynamics of time series features is strong.•We show good classification performance on the well-known PhysioBank database. Neurodegenerative diseases (NDDs), such as Parkinson’s disease (PD), Huntington’s disease (HD) and amyotrophic lateral sclerosis (ALS), create serious gait abnormalities. They lead to altered gait rhythm and gait dynamics which can be reflected by a time series of stride-to-stride measures of footfall contact times. The temporal fluctuations in gait dynamics provide us with a non-invasive technique to evaluate the effects of neurological impairments on gait and its variations with diseases. In this paper, we present a new method using gait dynamics to classify (diagnose) NDDs via deterministic learning theory. The classification approach consists of two phases: a training phase and a classification phase. In the training phase, gait features representing gait dynamics are derived from the time series of swing intervals and stance intervals of the left and right feet. Gait dynamics underlying gait patterns of healthy controls and NDDs subjects are locally accurately approximated by radial basis function (RBF) neural networks. The obtained knowledge of approximated gait dynamics is stored in constant RBF networks. Gait patterns of healthy controls and NDDs subjects constitute a training set. In the classification phase, a bank of dynamical estimators is constructed for all the training gait patterns. Prior knowledge of gait dynamics represented by the constant RBF networks is embedded in the estimators. By comparing the set of estimators with a test NDDs gait pattern to be classified, a set of test errors are generated. The average L1 norms of the errors are taken as the classification measure between the dynamics of the training gait patterns and the dynamics of the test NDDs gait pattern according to the smallest error principle. Finally, experiments are carried out to demonstrate that the proposed method can effectively separate the gait patterns between the groups of healthy controls and neurodegenerative patients.