A multicenter clinical AI system study for detection and diagnosis of focal liver lesions

Hanning Ying ¹, Xiaoqing Liu ², Min Zhang ³

¹Department of General Surgery, Sir Run Run Shaw Hospital, Zhejiang University School of Medicine, Hangzhou, China.
²Deepwise Artificial Intelligence Laboratory, Beijing, China.
³College of Computer Science and Technology, Zhejiang University, Hangzhou, China.
⁴School of Medicine, Zhejiang University, Hangzhou, China.
⁵Department of Radiology, Sir Run Run Shaw Hospital, Zhejiang University School of Medicine, Hangzhou, China.
⁶Zhangzhou Municipal Hospital of Fujian Province, Zhangzhou, China.
⁷Quzhou People's Hospital, Quzhou, China.
⁸Cancer Hospital of the University of Chinese Academy of Sciences (ZheJiang Cancer Hospital), Hangzhou, China.
⁹Zhejiang Provincial People's Hospital, Hangzhou, China.
¹⁰Department of Hepatopancreatobiliary Surgery, Ningbo First Hospital, Ningbo, China.
¹¹Hwa Mei Hospital, University of Chinese Academy of Sciences (Ningbo No.2 Hospital), Ningbo, China.
¹²Department of Hepatopancreatobiliary Surgery, Yinzhou People's Hospital, Ningbo, China.
¹³Department of Radiology, Huzhou Central Hospital, Affiliated Central Hospital of Huzhou University, Huzhou, China.
¹⁴Shaoxing People's Hospital, Shaoxing, China.
¹⁵The First Hospital of Jiaxing Affiliated Hospital of Jiaxing University, Jiaxing, China.
¹⁶The Second Hospital of Jiaxing Affiliated Hospital of Jiaxing University, Jiaxing, China.
¹⁷Jinhua Municipal Central Hospital, Jinhua, China.
¹⁸Jinhua GuangFU Hospital, Jinhua, China.
¹⁹Taizhou Municipal Central Hospital, Taizhou, China.
²⁰The First People's Hospital of Wenling, Taizhou, China.
²¹Lishui People's Hospital, Lishui, China.
²²Central Laboratory of Sir Run Run Shaw Hospital, Zhejiang University School of Medicine, Hangzhou, China.
²³Xiamen University, Xiamen, China.
²⁴Brigham and Women' Hospital, Harvard Medical School, Boston, MA, USA. xxu@bwh.harvard.edu.
²⁵Faculty of Medicine, Imperial College London, London, UK. l.jiao@imperial.ac.uk.
²⁶Department of Radiology, Second Affiliated Hospital of Zhejiang University School of Medicine, Hangzhou, China. risheng-yu@zju.edu.cn.
²⁷Faculty of Medicine, the Chinese University of Hong Kong, Hong Kong, China. josephlau@cuhk.edu.hk.
²⁸Department of Computer Science, The University of Hong Kong, Hong Kong, China. yizhouy@acm.org.
²⁹Department of General Surgery, Sir Run Run Shaw Hospital, Zhejiang University School of Medicine, Hangzhou, China. srrsh_cxj@zju.edu.cn.

⁰Department of General Surgery, Sir Run Run Shaw Hospital, Zhejiang University School of Medicine, Hangzhou, China.

Nature Communications +

|

February 7, 2024

View abstract on PubMed

Summary

A new AI system, LiAIDS, accurately diagnoses focal liver lesions, matching senior radiologists. It also improves human diagnosticians' performance and efficiently identifies low-risk patients, aiding routine clinical use.

Area Of Science

Medical Imaging
Artificial Intelligence
Hepatology

Background

Accurate diagnosis of focal liver lesions is critical for patient outcomes.
Current diagnostic methods can be variable, especially among less experienced clinicians.

Purpose Of The Study

To develop and validate a fully automated AI system (LiAIDS) for diagnosing focal liver lesions.
To assess LiAIDS's diagnostic performance compared to radiologists of varying experience levels.
To evaluate the impact of LiAIDS on radiologists' diagnostic accuracy.

Main Methods

Development and validation of the LiAIDS using a large dataset (12,610 patients) from 18 hospitals.
Retrospective and prospective study designs.
Comparative analysis of LiAIDS performance against junior and senior radiologists using F1-scores.
Triage study to assess LiAIDS's ability to classify patients by risk.

Main Results

LiAIDS achieved high F1-scores: 0.940 for benign and 0.692 for malignant lesions.
LiAIDS performance was comparable to senior radiologists and superior to junior radiologists.
Radiologists' diagnostic accuracy improved with LiAIDS assistance.
LiAIDS successfully triaged 76.46% of patients as low risk with 99.0% NPV.

Conclusions

LiAIDS demonstrates significant potential as a routine diagnostic tool for focal liver lesions.
The AI system can enhance the diagnostic capabilities of both junior and senior radiologists.
LiAIDS offers efficient risk stratification, aiding clinical workflow and patient management.

Related Concept Videos

Ultrasound II: Endoscopic Ultrasound and FibroScan

Endoscopic Ultrasound (EUS) and FibroScan are valuable diagnostic tools in gastroenterology and hepatology, each with specific applications and techniques.
Endoscopic Ultrasound (EUS):

Technological Integration: EUS is a sophisticated integration of optical endoscopy and ultrasonography. The endoscope visualizes the mucosal surface directly, while the ultrasound component penetrates deeper to visualize organs and structures beyond the gastrointestinal wall.
Image Formation: The ultrasound...