Control-group feature normalization for multivariate pattern analysis of structural MRI data using the support vector machine.

We just began a large new study on
Autism, ADHD, Anxiety & Depression in 12-17 year-olds.
Read more and learn how you can help!

TitleControl-group feature normalization for multivariate pattern analysis of structural MRI data using the support vector machine.
Publication TypeJournal Article
Year of Publication2016
AuthorsLinn, KA, Gaonkar, B, Satterthwaite, TD, Doshi, J, Davatzikos, C, Shinohara, RT
JournalNeuroimage
Volume132
Pagination157-166
Date Published2016 05 15
ISSN1095-9572
KeywordsAged, Alzheimer Disease, Brain, Brain Mapping, Computer Simulation, Female, Humans, Magnetic Resonance Imaging, Male, Multivariate Analysis, Pattern Recognition, Automated, Reproducibility of Results, Sensitivity and Specificity, Signal Processing, Computer-Assisted, Support Vector Machine
Abstract

Normalization of feature vector values is a common practice in machine learning. Generally, each feature value is standardized to the unit hypercube or by normalizing to zero mean and unit variance. Classification decisions based on support vector machines (SVMs) or by other methods are sensitive to the specific normalization used on the features. In the context of multivariate pattern analysis using neuroimaging data, standardization effectively up- and down-weights features based on their individual variability. Since the standard approach uses the entire data set to guide the normalization, it utilizes the total variability of these features. This total variation is inevitably dependent on the amount of marginal separation between groups. Thus, such a normalization may attenuate the separability of the data in high dimensional space. In this work we propose an alternate approach that uses an estimate of the control-group standard deviation to normalize features before training. We study our proposed approach in the context of group classification using structural MRI data. We show that control-based normalization leads to better reproducibility of estimated multivariate disease patterns and improves the classifier performance in many cases.

DOI10.1016/j.neuroimage.2016.02.044
Alternate JournalNeuroimage
PubMed ID26915498
PubMed Central IDPMC4851898
Grant ListK23 MH098130 / MH / NIMH NIH HHS / United States
R01 AG014971 / AG / NIA NIH HHS / United States
R01 MH107703 / MH / NIMH NIH HHS / United States
R01 NS085211 / NS / NINDS NIH HHS / United States
Comments
Leave a Comment