AI-Assisted Triage Using CT Head Scans For Emergency Diagnosis

Abstract

The main goal of CT-based triage is to shorten the time it takes to reach the expert opinion for patients in emergency situations, especially in cases of cranial fractures and intracranial hemorrhage. Increasing the performance in this regard may be possible with the use of artificial intelligence-supported software that may pre-scan the images and put them in order of urgency before the human triage officer is able to evaluate them. The project involves the development software that quickly classifies cases into one of two groups, urgent or non-urgent, by analysing the computerized tomography (CT) image of the brain taken without the administration of intravenous contrast material. In this way, more effective triage is aimed. The software developed for this purpose was observed to have high performances in two separate machine learning models. Additionally, a visual interface that allows viewing DICOM files was developed within the scope of the project.