Related Experiment Videos
Reasoning Model-Assisted Second-Reader Quality Control of Chinese-Language Ultrasound Reports: A Retrospective
Zhenqi Zhang1, Zirui Jiang2, Yihan Qi3
1General Affairs Office, Sichuan Clinical Research Center for Cancer, Sichuan Cancer Hospital & Institute, Sichuan Cancer Center, Affiliated Cancer Hospital of University of Electronic Science and Technology of China, Chengdu, 610041, China.
Journal of Imaging Informatics in Medicine
|June 29, 2026
Summary
The DeepSeek-R1 AI model shows promise as a second-reader for Chinese ultrasound reports, achieving high specificity for quality control. A combined AI-physician workflow significantly improved diagnostic accuracy.
Area of Science:
- Artificial Intelligence in Medical Imaging
- Natural Language Processing for Healthcare
- Diagnostic Accuracy Studies
Background:
- Quality control (QC) in medical reporting is crucial for patient safety.
- Automated tools can potentially enhance QC efficiency and accuracy.
- Chinese-language ultrasound reports require specialized validation methods.
Purpose of the Study:
- To assess DeepSeek-R1's capability as a second-reader QC tool for Chinese ultrasound reports.
- To compare the performance of DeepSeek-R1 against senior physicians in identifying report discrepancies.
- To evaluate the feasibility of integrating an AI model into the report finalization workflow.
Main Methods:
- Retrospective diagnostic-accuracy study involving 500 deidentified ultrasound reports.
- Parallel blinded review design comparing DeepSeek-R1 and physician reviewers.
- Independent evaluation of reports without access to other reviewers' outputs or the reference standard.
Main Results:
- DeepSeek-R1 demonstrated 69.1% sensitivity and 98.1% specificity.
- The AI model showed numerically higher sensitivity than senior physicians (69.1% vs. 47.1%).
- A combined AI-physician workflow simulation achieved 95.6% sensitivity and 96.3% specificity.
Conclusions:
- DeepSeek-R1 shows potential as a high-specificity second-reader for ultrasound report quality control.
- AI integration can enhance diagnostic accuracy and workflow efficiency in medical reporting.
- Human oversight remains essential for complex cases and final sign-off.
