Title: A classifier fusion strategy to improve the early detection of neurodegenerative diseases

Authors: Shamaila Iram; Paul Fergus; Dhiya Al-Jumeily; Abir Hussain; Martin Randles

Addresses: Applied Computing Research Group, School of Computing and Mathematical Sciences, Liverpool John Moores University, James Parsons Building, Byrom Street, L3 3AF, Liverpool, England, UK ' Applied Computing Research Group, School of Computing and Mathematical Sciences, Liverpool John Moores University, James Parsons Building, Byrom Street, L3 3AF, Liverpool, England, UK ' Applied Computing Research Group, School of Computing and Mathematical Sciences, Liverpool John Moores University, James Parsons Building, Byrom Street, L3 3AF, Liverpool, England, UK ' Applied Computing Research Group, School of Computing and Mathematical Sciences, Liverpool John Moores University, James Parsons Building, Byrom Street, L3 3AF, Liverpool, England, UK ' Applied Computing Research Group, School of Computing and Mathematical Sciences, Liverpool John Moores University, James Parsons Building, Byrom Street, L3 3AF, Liverpool, England, UK

Abstract: People in developed countries are living longer, and this has resulted in the prevalence of age-related diseases like Alzheimer's and dementia. Many believe that the early detection of neurodegenerative diseases will provide a much more sustainable framework for dealing with age-related diseases in the future. This paper considers this idea and proposes a new classifier fusion strategy that combines classification algorithms and rules (voting, product, mean, median, maximum and minimum) to measure specific behaviours in people suffering with neurodegenerative diseases. More specifically, the fusion strategy analyses the stride-to-stride intervals in gait and its correlation with neurological functions. This approach is compared with base level classifiers (a single classification algorithm) using a set of feature vectors associated with gait patterns obtained from neurodegenerative patients and healthy people. The results show that the fusion strategy improves classification. Our experiments successfully show that a fusion strategy generates better results and classifies subjects more accurately than base level classifiers.

Keywords: classifier fusion; pattern recognition; behaviour classification; machine learning; Huntington's disease; Parkinson's disease; amyotrophic lateral sclerosis; ALS; movement signals; artificial intelligence; early detection; disease detection; neurodegenerative diseases; age-related diseases; behaviour measurement; stride-to-stride intervals; gait patterns; neurological functions.

DOI: 10.1504/IJAISC.2015.067525

International Journal of Artificial Intelligence and Soft Computing, 2015 Vol.5 No.1, pp.23 - 44

Accepted: 10 Nov 2013
Published online: 19 Feb 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article