Abstract
Feature selection plays a key role in multi-voxel pattern analysis because functional magnetic resonance imaging data are typically noisy, sparse, and high-dimensional. Although the conventional evaluation criterion is the classification accuracy, selecting a stable feature set that is not sensitive to the variance in dataset may provide more scientific insights. In this study, we aim to investigate the stability of feature selection methods and test the stability-based feature selection scheme on two benchmark datasets. Top-k feature selection with a ranking score of mutual information and correlation, recursive feature elimination integrated with support vector machine, and L1 and L2-norm regularizations were adapted to a bootstrapped stability selection framework, and the selected algorithms were compared based on both accuracy and stability scores. The results indicate that regularization-based methods are generally more stable in StarPlus dataset, but in Haxby dataset they failed to perform as well as others.
Highlights
Exploring the mysteries of brain function is one of the most challenging and fascinating tasks in the domain of science
We would like to focus on multi-voxel pattern analysis (MVPA) [4], which is a commonly used methodological framework for analyzing functional magnetic resonance imaging data with machine learning algorithms. fMRI is a popular, non-invasive neuroimaging technique to measure brain activity via blood-oxygen-level dependent (BOLD) signals, recorded as time series in a three-dimensional (3D) brain space
We conducted a comprehensive analysis for a selection of filter, wrapper, and embedded feature selection approaches on the two benchmark fMRI datasets, adopting a stability-based methodological framework
Summary
Exploring the mysteries of brain function is one of the most challenging and fascinating tasks in the domain of science. With the aid of modern neuroimaging techniques, the capability of machine learning algorithms to identify distributed patterns of voxels in response to stimuli allows for decoding brain activities using data-driven models. We would like to focus on multi-voxel pattern analysis (MVPA) [4], which is a commonly used methodological framework for analyzing functional magnetic resonance imaging (fMRI) data with machine learning algorithms (see Fig. 1). FMRI is a popular, non-invasive neuroimaging technique to measure brain activity via blood-oxygen-level dependent (BOLD) signals, recorded as time series in a three-dimensional (3D) brain space. The precise spatial localization of brain activation, is an essential advantage of fMRI compared to other non-invasive neuroimaging techniques. MVPA constructs a pattern classification problem to decode neural information processing by characterizing multivariate brain activity patterns [5]
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have