Additional file 1 of RadCLARE: an automated clinical language engine for detecting semantic errors in radiology reports

Abstract

Additional file 1: Supplementary Table S1. Comprehensive definitions and categories of errors in radiology reports. Supplementary Table S2. Examples of commonly used phrases from imaging reports in the high-quality corpus used for error generation. Supplementary Table S3. Examples of homophone/near-homophone in the high-quality corpus used for error generation. Supplementary Table S4. Examples of error generation. Supplementary Table S5. Examples of Custom Entity Relationship Database. Supplementary Table S6. Subjective evaluations of RadCLARE in senior and attending radiologists. Supplementary Fig.S1. Illustration of the RadCLARE user interface for semantic error detection. Once the report has been finalised, the system automatically initiates error detection and sends error alerts back to the front end. The left panel displays a notification showing the number of errors detected. The errors in the report are highlighted in color in the right panel, along with correction suggetions. Supplementary Fig.S2. Line plots showing the semantic error rate of finalized reports per month. RadCLARE was implemented in June 2023 and the error rate of all groups decreased significantly. (a) Overall; (b) Different imaging modality; (c) Different time period for submission of the report (day: 8:00 AM to 5:59 PM; night: 6:00 AM to 7:59AM); (d) Different physician's title.

Ähnliche Arbeiten