The goal of this project is to demonstrate the feasibility of a computer aid for rapid and accurate diagnosis of aortic transection (AT) in the emergency department (ED). AT is a first concern in trauma cases involving motor vehicle impact. If present and not treated quickly, rapid death is the inevitable outcome. The definitive test for AT is an arteriogram. While essential for AT positive cases, this procedure has considerable morbidity, is time consuming, and expensive for AT negative cases. In Phase I, an artificial neural network (ANN) will be trained to predict AT from physician's findings from chest radiographs retrospectively obtained from trauma cases that were suspect for AT and for which arteriograms were obtained. Feasibility will be demonstrated if the ANN can decrease the false positive rate while maintaining overall accuracy. The final system will provide rapid, accurate, and consistent decision support for triage of trauma cases in the ED. Long term benefits include 1) improved patient care (reduction of critical time-consuming tests when unnecessary), 2) cost savings to patients, trauma centers, and third-party payers (reduction of expensive tests when unnecessary), and 3) potential for reduction in malpractice liability (documented consistency in decision making).Proposed commercial application:Approximately 100,000 vehicle impact cases with some suspicion of AT are received in the nation's emergency departments each year and receive an arteriogram at a cost of $2500. Our preliminary results indicate that about 60% of these arteriograms could have been avoided representing a potential savings of $ 150,000,000. To potentially save 60% of $2500 per case ($ 1500), AxIS anticipates that a charge of $500 per case is reasonable. This represents a potential annual projected income of $50,000,000 per year with an associated savings to the national health care system of $100,000,000 per year.National Institute of Heart, Lung, and Blood Institute (NHLBI)
