A genetic algorithm based feature selection for binary phenotype prediction using structural brain magnetic resonance imaging

Machine learning methods have been increasingly applied to structural brain magnetic resonance imaging (MRI) scans for predicting clinical phenotypes at the individual level. Despite significant methodological developments, reducing the dimensionality of the features extracted from brain MRI data remains a major challenge. In this paper, we propose a genetic algorithm based feature selection approach to binary phenotype prediction using structural brain MRI. We divide the population of individuals into multiple tribes and modify the initialization and evolutionary operations to ensure that the number of selected features in each tribe follows a Gaussian distribution. Thus each tribe is able to focus on exploring a specific part of the solution space. We also incorporate tribe competition into the evolution process, which allows the tribe that produces better individuals to enlarge its sizes so as to have more individuals to search the sub solution space it explores. We have evaluated our proposed approach against eight wrapper and nine filter feature selection methods on the binary phenotype prediction dataset used in the MICCAI 2014 Machine Learning Challenge. Our results indicate that the proposed approach can identify the optimal feature subset more effectively and is able to produce more accurate binary phenotype prediction.