Multivariate pattern analysis of brain structure: A diagnostic tool for Neurofibromatosis type 1

Abstract

Neurofibromatosis type 1 (NF1) is a genetic disorder characterized by increased predisposition for tumor development and cognitive deficits. In this work, we used maps of grey matter density obtained from Magnetic Resonance (MR) brain structural scans to distinguish between NF1 patients and healthy controls with a multivariate pattern analysis technique, Support Vector Machines. Up to 83% of all participants were correctly classified (mean sensitivity of 82%; mean specificity of 84%; significance level p < 0.01). This high level of classification accuracy of NF1 patients suggests this technique as a potential diagnostic tool. In addition, we determined the brain regions that the algorithm used to distinguish between NF1 patients and healthy controls. These regions were not identified as abnormal using univariate voxel-by-voxel comparison indicating that multivariate techniques are a useful powerful tool with which to identify potential structural defects in the NF1 brain.