Semi-Automated Radiological Report Generation

Every radiological analysis carried out requires a report to be written on that examination. Useful radiological reports need to include a lot of information. For instance, the Radiological Society of North America (RSNA) Radiology Reporting Committee's recommendations suggest that radiology reports must include the patient's clinical history, description of the imaging devices and technique, associated imaging parameters, description of salient findings and comparison to previous observations, among others. Consequently, authoring such detailed reports requires a lot of time and automating them, even partially, may free up radiologists to focus on other important and specialized tasks. 
In this project we aim to develop a controlled natural language designed to speed up the routine process of medical image interpretation — specifically the aspect or radiology report generation.

The projects' aims are to:

  • Create an international standard, based on published literature, of the information that ought to be included in radiology reports
  • Determine how current Zambian reports deviate from the created standard
  • Investigate whether any deviation, if found in Zambian reports, is by design and preferred by practising radiologists
  • Create a controlled natural language using Grammatical Framework (https://www.grammaticalframework.org/) that codifies the needs of practising Zambian radiologists based on the created standard and the preferences of Zambian radiologists.
  • Create a human-friendly interface for the controlled natural language that can be used by practising Zambian radiologists to author reports
  • Investigate the impact of the radiology report authoring tool as a function of authoring time and quality of radiology reports

Project Members

Academic Staff
University Teaching Hospitals
Academic Staff
University of Cape Town
Academic Staff
University Teaching Hospitals

Publications

2024
Mahlaza, Zola, Ernest Obbie Zulu, and Lighton Phiri. 2024. “Radiology Report Terminology To Characterise Reports In Southern Africa”. In 17Th International Conference On Metadata And Semantics Research. Milan, Italy: Springer Nature.
View Publication