Traumatic brain injury (TBI) is still the most common cause of death and long-term disability due to nervous system damage in the world. One of the main reasons for this is the drastic increase in intracranial pressure (ICP) in cases of moderate to severe TBI. The traditional method of controlling ICP involves continuous monitoring of pre-established threshold values and administering treatments only when necessary. However, this strategy often fails to grasp the intricate, non-linear and time-related physiological changes that eventually lead to increased intracranial pressure. In this work, an AI-based approach together with a machine learning technique is proposed for the crisis detection of ICP in TBI patients, which would be entirely based on high-resolution physiological data and clinical variables. The research was retrospective and observational, and it was performed in MIMIC-III and MIMIC-IV patients' ICU databases. It consisted of adult ICU patients with TBI that were subjected to invasive ICP monitoring. Continuous ICP and blood pressure waveforms were combined with demographic, neurological, and clinical intervention data. ICP crisis was determined by a rise of ≥20–22 mmHg for at least five minutes, and a labeling technique involving sliding time-window was applied. A variety of the supervised ML models such as logistic regression, random forests, support vector machines, gradient boosting, RNNs, and LSTMs, etc., were created and tested through patient-wise data division to ensure no mixing of information occurs. The performance of the models was evaluated using clinically significant measures like area under the receiver operating characteristic curve (AUROC), sensitivity, specificity, precision, recall, and early warning time. Classic ML models showed strong predictive power on developed features, while LSTM-based models got better results by recognizing long term time dependencies in raw waveform data which made earlier and more accurate detection of coming ICP crises possible. The results indicate that the use of AI in predictive modeling can greatly improve the early risk assessment and the support of proactive interventions in neurocritical care. This framework highlights the potential of machine learning based decision support systems to transition ICP management from reactive threshold-based approaches toward anticipatory, personalized, and precision-guided care for patients with traumatic brain injury
Keywords: Traumatic Brain Injury; Intracranial Pressure; Machine Learning; Artificial Intelligence; Neurocritical Care; Time-Series Prediction
How to cite this article:Saeed S, Riaz M, Qadeer M, Riaz I, Shah HK..; Artificial intelligence and machine learning for predicting intracranial pressure crises in TBI patients..Int J Drug Deliv Technol. 2026;16(1s): 714-726; DOI: 10.25258/ijddt.16. 714-726