Web-based calculator using machine learning to predict intracranial hematoma in geriatric traumatic brain injury

Background: Traumatic brain injury (TBI) is a significant contributor to mortality and impairment among the general population. The elderly are at a higher risk of developing cerebral hematomas following TBI. Therefore, there has been an overuse of cranial computed tomography (CT) in this group. The purpose of this study was to assess the predictive ability of machine learning (ML) algorithms for traumatic intracranial hematoma prediction. The secondary objective was to explore the predictors associated with positive CT scans. Methods: A retrospective cohort study was conducted to examine TBI patients aged 60 years and older. To train the ML models, 70% of the data was separated, with the remaining 30% being used for testing. The supervised techniques used for training the ML models were na & iuml;ve Bayes (NB), support vector machines (SVM), k -nearest neighbor (KNN), decision trees (DT), random forests (RF), artificial neural networks (ANN), and extreme gradient boosting (XGB). Therefore, the testing dataset was used to evaluate the ML models' prediction capabilities. Results: There were 2,052 patients in the total cohort and 403 (19.6%) of the cohort had positive CT scans. Ten clinical predictors were used for building ML models and testing their performance. The NB algorithm had acceptable discrimination; the area under the receiver operating characteristic curve (AUC) was 0.70. Moreover, the sensitivity and F1 score of NB were 0.97 and 0.91, respectively. Conclusions: ML models have the potential to serve as a screening tool for predicting positive cranial CT scans in elderly TBI patients since they can assist clinicians in making clinical decisions. In practice, a web application would be a simple way to apply the predictive ML model. Furthermore, future studies should involve external validation to examine the generalizability of clinical prediction systems.