Classification of traumatic brain injury using support vector machine analysis of event-related Tsallis entropy

An estimated 1.4 million Americans suffer from traumatic brain injury (TBI) each year. Current methods of detecting TBI, such as computerized tomography (CT), magnetic resonance imaging (MRI), and Positron Emission Tomography (PET) scanning are time-consuming and expensive. Here, the viability of a potentially more cost-effective means of detecting TBI is presented. Support vector machine (SVM) analyses are employed to classify 15 TBI and 15 normal individuals´ EEG recordings taken during a working memory test. The features used by the SVM analyses include different sets of event-related Tsallis entropy functionals. The analyses demonstrate a strong correlation between the event-related functionals (ERFs) and the presence of TBI, attaining classification accuracies as high as 90%.