Authors
Shan-Shan Hu, Ze-Yu Cao, Mi Ning, Xin Xiao, Xian-Min Deng, Hong-Ze Zeng, Chan Qiu, Li Zhong, Hang Yi, Ai-Min Liu, Dan-Ping Huang, Zhi-Hang Zhou, Bo Ning
Published in
Digestive and liver disease : official journal of the Italian Society of Gastroenterology and the Italian Association for the Study of the Liver. Aug 14, 2026. Epub Aug 14, 2026.
Abstract
Gastrointestinal stromal tumors (GISTs) are tumors with malignant potential. This research aims to develop an artificial intelligence (AI)-based system for analyzing endoscopic ultrasonography (EUS) visuals and generating risk scores.
This system is designed to better predict GIST risk levels by identifying risk factors.
An internal dataset comprising 504 EUS images from 226 patients with pathologically confirmed GISTs was collected from Yuzhong Hospital and Jiangnan Hospital of the Second Affiliated Hospital of Chongqing Medical University, Sichuan Provincial People's Hospital between 2018 and 2024. Multiple AI-based machine learning models were developed and tested. Six machine learning models were compared, and the optimal diagnostic model was selected based on statistical results.
For the best-performing model, XGBoost, the internal test set results were as follows: overall accuracy 83.17%, sensitivity 75.68%, specificity 93.72%, positive predictive value (PPV) 73.21%, negative predictive value (NPV) 93.52%, F1 score 0.74, and area under the curve (AUC) 0.97. The external validation set results were: overall accuracy 80.68%, sensitivity 71.53%, specificity 92.72%, PPV 70.44%, NPV 92.78%, F1 score 0.70, and AUC 0.94.
This model utilizes machine learning algorithms on EUS images to accurately predict GIST risk stratification.
PMID:
42601260
Bibliographic data and abstract were imported from PubMed on 15 Aug 2026.
Read full publication at:
Please sign in
to see all details.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 0
- Comments 0