Artificial Intelligence-Aided Head Computed Tomography Analysis to Predict Outcome in Civilian Firearm-Inflicted Brain Injury
Abstract
Civilian firearm-inflicted brain injury (CFI-BI) is of global epidemiologic concern. Guidelines on management of CFI-BI are outdated and offer little guidance for management or prognostication of this disease state. Current models used to predict outcome after gunshot wounds to the head are complicated and often times confounded by concomitant injury as well as the compromised overall clinical state of the patient. This disease state, while prevalent in large cities within the United States, is a hallmark of military trauma. Still, what little literature is available regarding outcomes and management is extrapolated from military data. At the University of Chicago Medical Center, we treat high rates of victims of gunshot wounds, averaging at over 250 victims over the last 3 years (>1/week) with ~ 50% survival rate, half of which have are in a good functional state at the time of discharge. This rate of gunshot wounds to the head compares to a handful of trauma centers across the world. Over the last 3 years, in addition to caring for patients, we have published multiple peer-reviewed manuscripts on the topic, including the role of coagulopathy and vascular injury in this patient population. We believe that poorly understood and peculiar mechanisms govern severity of injury and outcomes in victims of gunshot wounds to the head. The mechanism of injury in CFI-BI is largely structurally attributable to direct and indirect injuries from the bullet passing through brain tissue. The prevailing imaging modality at this time is computed head tomography (HCT) and remains limited by image artifact caused by the bullet as well as resolution restrictions that limit the ability to assess the full extent of the presumed injury. Our group is the primary team that cares for CFI-BI in the neurosciences ICU at the University of Chicago. Recently, in collaboration with colleagues from medical physics and computer science, we were able to develop an algorithm that predicts onset of brain injury in survivors of cardiac arrest using images that were read as being normal by expert neuro-radiologists and that were obtained a few hours after resuscitation from cardiac arrest. Therefore, and seeing as injury to the brain in the context of gunshots to the head is predominantly structural, we propose to leverage the high volume of CFI-BI patients we treat as well as our successful interdepartmental collaboration to utilize artificial intelligence/computer vision to analyze early HCT in CFI-BI. We will predict patient outcome irrespective of other clinical characteristics at the time. We will also explore the model to identify the salient imaging features that it uses to classify outcome. In doing so, we will then examine the association of those features with the patient’s clinical state and laboratory findings as well as the temporal progression of those features. We hope to describe novel and unique imaging features that help us understand the kind of damage gunshot wounds cause in the brain and ultimately identify new therapeutic targets specific to this type of brain injury. This work is of particular relevance to the military context since gunshot wounds to the head are a prevalent cause of morbidity in that setting. The model described will allow for more efficient triage and resource allocation in combat zones. It will also identify unique disease specific features that may be used to develop treatment strategies for these patients.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Dec 28, 2022
- Source ID
- W81XWH2210480
Entities
People
- Ali Mansour
Organizations
- United States Army
- University of Chicago