Traumatic Brain Injury (TBI) is a complex condition that often results in long-term cognitive, emotional, and physical impairments. This study proposes a deep learning-based model utilizing MRI data to predict TBI severity. The model uses a residual learning convolutional neural network (CNN), which leverages transfer learning to reduce training time while enhancing predictive accuracy. The dataset consists of MRI brain scans from 204 TBI patients, categorized based on severity levels. We used key metrics such as accuracy, sensitivity, specificity, and area under the receiver operating characteristic (AUC-ROC) curve to evaluate the model’s performance. The model achieved an accuracy of 93.31%, with high sensitivity for severe TBI cases (100%) and slightly lower sensitivity for mild cases (78.32%). These results demonstrate the potential clinical applicability of the proposed model in improving early diagnosis and severity assessment of TBI. Future research will expand the dataset and refine the model to enhance its robustness and generalizability.