Abstract
Language dominance behavior is identified as typical/atypical based on the asymmetry of the brain activation. Typical language is commonly defined by left brain hemisphere activation dominance during performing language tasks, while atypical language involves either right or both brain hemispheres. Traditionally, the methods used to identify the asymmetry of the brain activation are expert visual assessment and lateralization index (LI) computation. This paper presents a novel application of a supervised learning machine paradigm called Nonlinear Decision Functions (NDF). The merits of this paradigm are exploited on providing an automatic procedure for the identification of typical/atypical language dominance. NDF are invaluable tools for the resolution of real-world problems such as the one addressed in this paper. To identify language behavior, the subject undergoes an fMRI test. The resulting 4-D dataset (3D spatial information plus time series) is processed. Based on statistical and image analyses, a brain activation map (BAM) is generated. A total of 103 fMRI datasets from 5 different hospitals were analyzed, with 64 healthy control (HC) datasets, and 39 LRE datasets. On using NDFs on the basis of the demographics as well as the extent and intensity of these BAMs, the results obtained yielded a sensitivity of 80.6%, a specificity of 70.5%, an accuracy of 97.8% and a precision of 98.2%.