Keyword-based AI assistance in the generation of radiology reports: A pilot study.
Authors
Affiliations (4)
Affiliations (4)
- Department of Radiology, the Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, Zhejiang, 310009, China.
- Department of Radiology, Changxing people's hospital, No. 66 Taihu Lake Middle Road, Huzhou, 313100, China.
- Department of Radiology, Yuyao People's Hospital of Zhejiang Province, the Affiliated Yangming Hospital of Ningbo University, Yuyao, Zhejiang, 315400, China.
- Department of Radiology, the Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, Zhejiang, 310009, China. [email protected].
Abstract
Radiology reporting is a time-intensive process, and artificial intelligence (AI) shows potential for textual processing in radiology reporting. In this study, we proposed a keyword-based AI-assisted radiology reporting paradigm and evaluated its potential for clinical implementation. Using MRI data from 100 patients with intracranial tumors, two radiology residents independently wrote both a routine complete report (routine report) and a keyword report for each patient. Based on the keyword reports and a designed prompt, AI-assisted reports were generated (AI-generated reports). The results demonstrated median reporting time reduction ratios of 27.1% and 28.8% (mean, 28.0%) for the two residents, with no significant difference in quality scores between AI-generated and routine reports (p > 0.50). AI-generated reports showed primary diagnosis accuracies of 68.0% (Resident 1) and 76.0% (Resident 2) (mean, 72.0%). These findings suggest that the keyword-based AI-assisted reporting paradigm exhibits significant potential for clinical translation.