Friday 31 January 2025
A team of researchers has developed a new approach to evaluating the quality of automated radiology reports, which can help doctors and hospitals improve patient care by ensuring that medical imaging results are accurately reported.
The traditional method of evaluating radiology reports is based on lexical or textual semantics, where the report’s content is compared to a set of pre-defined criteria. However, this approach has limitations, as it does not take into account the fine-grained details of the report, such as the location and severity of findings.
To address this issue, the researchers developed a new method that extracts fine-grained finding patterns from radiology reports, which capture the presence or absence of specific findings, along with their locations, laterality, and severity. These patterns are then compared to anatomical locations on chest X-ray images to evaluate the report’s quality.
The team tested their approach using a dataset of 439 chest X-ray images with validated ground truth reports and extracted FFL patterns covering 60 findings. They evaluated three different report generators – RGRG, XrayGPT, and GPT4 – using their new method and compared the results to other evaluation scores.
The results showed that the RGRG report generator had the highest lexical quality, followed closely by the GPT4 generator. The team’s approach also demonstrated good sensitivity to factual errors in radiology reports, which is an important feature for fact-checking of medical imaging reports.
In addition to improving the accuracy of automated radiology reports, this new approach can also help reduce the workload of radiologists and improve patient care by providing a more comprehensive evaluation of report quality.
Cite this article: “Enhancing Radiology Report Quality Through Fine-Grained Finding Patterns”, The Science Archive, 2025.
Automated Radiology Reports, Quality Evaluation, Lexical Semantics, Fine-Grained Finding Patterns, Chest X-Ray Images, Anatomical Locations, Report Generators, Sensitivity, Factual Errors, Patient Care.
