AI enhanced diagnostic accuracy and workload reduction in hepatocellular carcinoma screening
Rui-Fang Lu1, Chao-Yin She2, Dan-Ni He3
1Department of Medical Ultrasonics, Institute of Diagnostic and Interventional Ultrasound, MedAI Collaborative Lab, Ultrasomics Artificial Intelligence X-Lab, The First Affiliated Hospital of Sun Yat-sen University, Guangzhou, China.
NPJ Digital Medicine
|August 2, 2025
Summary
Artificial intelligence (AI) improves hepatocellular carcinoma (HCC) screening by enhancing ultrasound accuracy and reducing radiologist workload. Strategy 4, a human-AI collaboration, demonstrated superior performance in detection and classification.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Medicine
- Hepatology
Background:
- Hepatocellular carcinoma (HCC) screening via ultrasound faces challenges in accuracy and radiologist workload.
- Existing AI tools require optimization for clinical integration.
Purpose of the Study:
- To evaluate AI-enhanced strategies for improving HCC ultrasound screening accuracy.
- To assess the impact of AI on radiologist workload during HCC screening.
Main Methods:
- Retrospective, multicenter study of 21,934 liver ultrasound images from 11,960 patients.
- Assessed four AI-enhanced strategies using UniMatch for detection and LivNet for classification.
- Compared AI strategies against traditional radiologist evaluation.
Main Results:
- Strategy 4, combining AI detection with radiologist review of negative cases, showed improved specificity (0.787 vs. 0.698) and comparable sensitivity (0.956 vs. 0.991) to the original algorithm.
- Reduced radiologist workload by 54.5% and decreased recall and false positive rates.
- Demonstrated a successful human-AI collaboration model.
Conclusions:
- AI-enhanced HCC ultrasound screening can significantly improve accuracy and efficiency.
- Human-AI collaboration offers a promising approach to mitigate radiologist workload and enhance patient care.
- This strategy reduces unnecessary patient anxiety and system burden.
