Deep learning models for triaging hospital head MRI examinations

• DOI: 10.1016/j.media.2022.102391
• Publication Date: 2022-02-12T08:15:04Z
• Authors: David A. Wood; Sina Kafiabadi; Ayisha Al Busaidi; Emily Guilhem; Antanas Montvila; Jeremy Lynch; Matthew Townend; Siddharth Agarwal; Asif Mazumder; Gareth J. Barker; Sebastien Ourselin; James H. Cole; Thomas C. Booth
• Keywords: Triage; Flagging; Neuroradiology

The growing demand for head magnetic resonance imaging (MRI) examinations, along with a global shortage of radiologists, has led to an increase in the time taken report MRI scans recent years. For many neurological conditions, this delay can result poorer patient outcomes and inflated healthcare costs. Potentially, computer vision models could help reduce reporting times abnormal examinations by flagging abnormalities at imaging, allowing radiology departments prioritise limited resources into these first. To date, however, difficulty obtaining large, clinically-representative labelled datasets been bottleneck model development. In work, we present deep learning framework, based on convolutional neural networks, detecting clinically-relevant minimally processed, hospital-grade axial T2-weighted diffusion-weighted scans. were trained scale using Transformer-based neuroradiology classifier generate dataset 70,206 from two large UK hospital demonstrate fast (< 5 s), accurate (area under receiver operating characteristic curve (AUC) > 0.9), interpretable classification, good generalisability between hospitals (ΔAUC ≤ 0.02). Through simulation study show that our best would mean 28 days 14 9 demonstrating feasibility use clinical triage environment.