Detection of prenatal alcohol exposure using machine learning classification of resting-state functional network connectivity data

Summary

In this project, colleagues and I were interested in learning whether or not a machine learning technique could be used to discriminate between brain connectivity patterns associated with prenatal alcohol exposure from those of controls. We utilized a rodent model of prenatal alcohol exposure, resting state function magnetic resonance imaging (fMR), and group independent components analysis to measure functional connectivity. The functional connectivity measures were used in a leave one out cross validation procedure incorporating multiple SVM kernels for binary classification. We discovered that a quadratic SVM kernel was approximately 80% accurate in correctly classifying brain connectivity patterns. This research suggests that this approach may yield viable non-invasive diagnostic methods for fetal alcohol spectrum disorder with future refinements of the technique.