Trust-Enhanced Natural Processing for Standardisation of Radiology Reports

The first aim is to develop NLP system that can transform ASR-transcribed unstructured reports from radiologists into structured ones automatically, acting as a middleman between the two parties of the radiologist and the referring clinician. This system would allow for radiologists to continue to work in the way that they are used to but would provide referring clinicians reading the reports to receive information in a more reliable way. This would also provide interesting avenues for electronic health records and archiving healthcare data in a way for better analysis.

By implementing local language modelling for the domain and clinician, and automatic information extraction and analysis, we could harness this information and use it to further pedagogical understanding and practice to improve feedback for NIAW trainees, and result in higher quality and more consistent care administered to patients.