"Automated Decision Support System for Traumatic Injuries"
Abstract
Trauma is the leading cause of death among Americans younger than 46. Traumatic brain injuries and abdominal trauma contribute to most of this traumatic mortality and morbidity. Traumatic brain injury can cause a number of difficulties such as hemorrhage, shift in the brain structure (e.g., midline shift), and traumatic abdominal injury can lead to multiple complications including laceration of major organs such as liver, kidneys, and spleen. One of the most important initial tests performed on trauma patients to evaluate the severity of injury is computed tomography (CT) imaging. It provides critical information which is found to be a significant predictor of trauma severity and outcome.
The overall objective of this research is to develop a computer-aided decision support system for polytrauma. This system consists of two general components: one that performs image processing, while the other uses machine learning for outcome prediction. Here I will present image processing algorithms which automatically segment subdural hematoma, a key abnormality in the brain, as well as the liver and kidneys in the abdomen, using CT scans. Furthermore, in the outcome prediction phase, patient-level data from electronic health records is integrated with the features extracted from the CT scans to develop predictive models that support and enhance clinical decision-making.
