TBI is a biologically very complex and heterogenous condition. Important factors such as injury severity or type, age and basic clinical parameters observed immediately after injury are important in determining mortality, but the severity of morbidity cannot currently be assessed. In order to test our hypothesis that there are individuals more resilient to TBI, we need to develop an advanced prediction model that tests important parameters collected at admission as well as during the intensive care that might explain why some individuals believed to have similar injuries recover differently. Multiple prognostic models for TBI have accumulated over the last decades but none of them are widely used in clinical practice. While there are many prognostic models, they typically have several limitations such as small sample size, no external validation, and limited clinical user friendliness, furthermore, they only predict mortality a dichotomise fashion and a short time after the injury has occurred. A dichotomized outcome prediction tool limits the ability to stratify patients for clinical studies. A more advanced tool can be used for targeted recruitment to clinical treatment trials and tracking of any treatments that are undergoing clinical trial both in terms of efficacy of treatment, but also for safety monitoring. In our model TBI-POP, we used the static admission variable used in IMPACT model and performed external validations on a TBI cohort from Leuven, Belgium and the PROTECTIII study from Atlanta, USA. In order to advance the model and its accuracy longitudinal physiological data obtained by monitoring the injured brain i.e. intracranial pressure (ICP), cerebral perfusion pressure (CPP) and cerebral autoregulation (PRx) will be added. In addition, we will use an automated segmentation model for CT-brain scans developed for Alzheimer’s disease using AI. The assessment provided by this AI model (U-Net deep learning model) previously required MRI scanning, but it is now compatible with CT-scan. A CT-brain scan is the golden standard to image the injured brain and is always performed upon admission of a TBI patient. These scans are usually used by visual assessment of brain integrity and exclusion of co-pathologies. However, this AI model, a deep learning-based segmentation network can classify brain tissue in more detail and address the heterogeneity of underlying pathology almost equivalent to MRI scanning. Thus, we aim to add these parameters to our prognostic calculations and assess outcome. As a contingency plan, if the CT-AI- model cannot be incorporated with the prediction model, the evaluations of the CT-scans already assessed by neuroradiologists will be used as previously. This will move us toward data-driven care of TBI patients, which is about drawing scientific conclusions and making care recommendations on the basis of vast datasets, and about using large amounts of data to manage healthcare.