Automatic anomaly detection for brain imaging triage and classification (360G-Wellcome-103709_Z_14_A)

Modern brain imaging contains vastly more information than historical radiographs, yet its clinically informative output has remained the same: a radiologist’s verbal report. As the information content of imaging increases, a void has opened between what we expensively collect and what we actually use.This is both a lost opportunity, and an obstacle to the continued growth of brain imaging. Technology being developed by Dr Parashkev Nachev and colleagues at University College London seeks to close this gap by applying novel computer-assisted algorithms so as to exploit much more of the information in each scan than a verbal report contains. An automatic “anomaly map” for each scan, indexing the deviation from normality of each point, will assist radiological reporting, allow the application of computer systems that predict clinical outcomes from patterns of anomaly, and guide radiological triage and resource/performance management. The project aims to demonstrate the feasibility, robustness, clinical, and managerial value of the approach using a large collection of standard brain imaging, and to deliver a pilot system capable of translation into a full clinical product. Without changing any clinical pathways or adding new investigations, the system will improve radiological reporting and optimise radiological triage and management, while creating a scalable major new platform for computational imaging analysis.