Traumatic Brain Injury (TBI) is a significant public health concern, causing enduring neurological deficits that deeply impact patient and family quality of life, along with considerable socioeconomic burdens. These sequelae often encompass cognitive, behavioral, memory, attention, and mood disturbances, frequently leading to permanent disabilities. This study presents an Artificial Intelligence (AI)-driven strategy for TBI detection, using radiomic analysis of Magnetic Resonance Imaging (MRI) scans. The approach involves, first, a preprocessing phase to standardize MRI data, mitigating variability from diverse acquisition parameters. Second, Regions Of Interest (ROIs) are segmented using the SRI24 brain atlas, enabling the extraction of statistical and textural features to characterize brain structure. Finally, these features are used to train supervised learning models for TBI classification. Experimental results show that Logistic regression has the best performance (AUC = 1.00). Besides, identified patterns in the frontal cortex, brainstem, thalamus, and the amygdala-hippocampus complex suggest these regions are potential biomarkers for TBI.