Computed tomography (CT) structured reporting generation based on the intelligent assisted diagnosis system for...

To evaluate the clinical utility of the intelligent assisted diagnosis system for rectal cancer staging (IADS-R) in aiding radiologists of varying levels of experience to produce computed tomography (CT) structured reporting (SR).

Six hundred fifty abdominal enhanced CT datasets from rectal cancer patients were collected, with tumour and lymph node regions annotated. Four hundred images were used to train IADS-R's convolutional neural network model and 250 images were used for validation. By comparing the consistency and accuracy of diagnostic staging, the consistency of SR measurements, and the time efficiency in report generation among three groups of physicians (senior radiologists, junior radiologists, and junior radiologists assisted by IADS-R), the study demonstrated the supportive role of IADS-R for junior radiologists.

Junior radiologists with IADS-R assistance achieved diagnostic accuracy rates of 81.2% for T-staging and 81.6% for N-staging, comparable to senior radiologists (78.0% for T and 77.6% for N) and significantly higher than unassisted juniors (60.8% for T and 55.2% for N). SR generated by junior radiologists with IADS-R assistance were more consistent with senior radiologists' reports, particularly in measuring tumour size and proximity to pelvic tissue structures. Report generation time for IADS-R-assisted juniors (1773.1s) matched seniors and was faster than unassisted juniors (2356.1s). With the assistance of IADS-R, the efficiency and accuracy of junior radiologists in producing SR have been significantly improved.

IADS-R significantly improved junior radiologists' accuracy and efficiency in rectal cancer staging and SR generation, reaching levels comparable to senior radiologists.