Content area
Full text
About the Authors:
Margarita Kotti
* E-mail: [email protected]
Affiliations Biodynamics Laboratory, Division of Surgery, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London, London, United Kingdom, Brain Behaviour Laboratory, Department of Bioengineering, Imperial College London, London, United Kingdom
Lynsey D. Duffell
Affiliation: Biodynamics Laboratory, Division of Surgery, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London, London, United Kingdom
Aldo A. Faisal
Affiliations Brain Behaviour Laboratory, Department of Bioengineering, Imperial College London, London, United Kingdom, Department of Computing, Imperial College London, London, United Kingdom, MRC Clinical Sciences Centre, Faculty of Medicine, Imperial College London, London, United Kingdom
Alison H. McGregor
Affiliation: Biodynamics Laboratory, Division of Surgery, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London, London, United Kingdom
Introduction
The aim of this study is to check whether the redundant dimensionality of the human biomechanical system can be effectively reduced by projection into a low principal component (PC) space. In the later space it is proven that the patterns produced by normal subjects and pathological subjects that suffer from knee osteoarthritis (OA), are still identifiable. A challenge in analysing gait patterns is that, as a form of behaviour, it exhibits high variability [1]. Both sensory inputs and motor outputs are subjected to noise and uncertainty [2] [3].
Additionally, movement analysis is extremely complex since the musculoskeletal system has over 600 degrees of freedom. We assume that the design of our muscoloskeletal system is redundant and, as a result of, this the central nervous system has several options when generating movement for a specific task. In [4] it is indicated that muscular redundancy is necessary, however the idea of redundancy still greatly increases the complexity incurred when generating movement. Movement data is inherently variable both within subjects (across trials) as well as across subjects [5]. Most traditional motion analysis methods simply average away the variability in the data to obtain a clear readout of an underlying mechanism. This dismisses a lot of the obtained data implying that features buried in the structure of variability of behavioural data are lost. In contrast, we embrace here the variability of the data, as we hypothesis the structure of variability provides insight into the underlying mechanisms. The novelty of our idea is that instead...