Dataset Information

Artificial Intelligence in the Prediction of Gastrointestinal Stromal Tumors on Endoscopic Ultrasonography Images: Development, Validation and Comparison with Endosonographers.

ABSTRACT:

Background/aims

The accuracy of endosonographers in diagnosing gastric subepithelial lesions (SELs) using endoscopic ultrasonography (EUS) is influenced by experience and subjectivity. Artificial intelligence (AI) has achieved remarkable development in this field. This study aimed to develop an AI-based EUS diagnostic model for the diagnosis of SELs, and evaluated its efficacy with external validation.

Methods

We developed the EUS-AI model with ResNeSt50 using EUS images from two hospitals to predict the histopathology of the gastric SELs originating from muscularis propria. The diagnostic performance of the model was also validated using EUS images obtained from four other hospitals.

Results

A total of 2,057 images from 367 patients (375 SELs) were chosen to build the models, and 914 images from 106 patients (108 SELs) were chosen for external validation. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of the model for differentiating gastrointestinal stromal tumors (GISTs) and non-GISTs in the external validation sets by images were 82.01%, 68.22%, 86.77%, 59.86%, and 78.12%, respectively. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy in the external validation set by tumors were 83.75%, 71.43%, 89.33%, 60.61%, and 80.56%, respectively. The EUS-AI model showed better performance (especially specificity) than some endosonographers. The model helped improve the sensitivity, specificity, and accuracy of certain endosonographers.

Conclusions

We developed an EUS-AI model to classify gastric SELs originating from muscularis propria into GISTs and non-GISTs with good accuracy. The model may help improve the diagnostic performance of endosonographers. Further work is required to develop a multi-modal EUS-AI system.

SUBMITTER: Lu Y

PROVIDER: S-EPMC10651383 | biostudies-literature | 2023 Jan

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Artificial Intelligence in the Prediction of Gastrointestinal Stromal Tumors on Endoscopic Ultrasonography Images: Development, Validation and Comparison with Endosonographers.

Lu Yi Y Wu Jiachuan J Hu Minhui M Zhong Qinghua Q Er Limian L Shi Huihui H Cheng Weihui W Chen Ke K Liu Yuan Y Qiu Bingfeng B Xu Qiancheng Q Lai Guangshun G Wang Yufeng Y Luo Yuxuan Y Mu Jinbao J Zhang Wenjie W Zhi Min M Sun Jiachen J

Gut and liver 20230126 6

<h4>Background/aims</h4>The accuracy of endosonographers in diagnosing gastric subepithelial lesions (SELs) using endoscopic ultrasonography (EUS) is influenced by experience and subjectivity. Artificial intelligence (AI) has achieved remarkable development in this field. This study aimed to develop an AI-based EUS diagnostic model for the diagnosis of SELs, and evaluated its efficacy with external validation.<h4>Methods</h4>We developed the EUS-AI model with ResNeSt50 using EUS images from two ...[more]

PMID: 36700302

Similar Datasets

Project description:BackgroundThis retrospective study aimed to evaluate the factors influencing the accuracy of Endoscopic Ultrasonography (EUS) as a preoperative assessment for gastrointestinal tumors.MethodsA total of 261 patients with 264 gastrointestinal tumors were enrolled in the study. The parameters of the gastrointestinal lesions examined under EUS and their pathology were recorded and analyzed.ResultsThe accuracy of EUS for detecting intramucosal lesions and subepithelial lesions (SELs) were 83.6% and 91.4%, respectively. One hundred and ninety-four (73.5%) lesions originated from the mucous layer, as determined by pre-operation EUS examinations. The accuracy of EUS in predicting the correct T stage for intramucosal lesions in the gastric region, esophagus, and colorectum was 77%, 71.8%, and 84.6%, respectively. According to the Paris endoscopic classification, the distribution of macroscopic patterns was different between the EUS-pathology conformity and nonconformity groups (p = 0.018). In the nonconformity group, 48.6% of erosive lesions were classified as 0-IIc, 0-IIa + IIc, 0-IIc + IIa or 0-III macroscopic patterns compared with 26% patients in the conformity group (p = 0.025). Univariate analyses demonstrated that ulcerative lesions (OR = 7.516, 95% Confidence Interval [CI] 2.574-21.952, p < 0.001), location at the cardia of the stomach (OR = 3.619, 95%CI 1.076-12.168, p = 0.038), malignant tumor (OR = 2.920, 95%CI 1.339-6.368, p = 0.007) were significantly associated with EUS inaccuracy. Multivariate logistic regression analyses showed that ulcer was an independent risk factor associated with EUS inaccuracy, with odds ratios of 5.094 (95% CI: 1.641-15.807, p = 0.005).ConclusionsOur findings suggested that EUS is a reliable and easy-to-use diagnostic tool in decision-making regarding appropriate endoscopic treatment for gastrointestinal tumors. However, the diagnostic accuracy of EUS appeared questionable in the presence of ulceration.

Project description:BackgroundDespite the recent large number of studies comparing endoscopic and laparoscopic resection for small gastrointestinal stromal tumors (GISTs) (diameter ≤ 5 cm), the results remain conflicting. The objective of this work was to perform a cumulative meta-analysis to assess the advantages and disadvantages of endoscopic resection vs. laparoscopic resection.MethodsThe meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. We searched medical databases up to January 2020. Meta-analytical random or fixed effects models were used in pooled analyses. Meta-regression, cumulative meta-analyses, and subgroup analyses were performed to improve the accuracy of the conclusion. Sensitivity analyses were applied to assess the robustness of the results.ResultsA total of 12 cohort studies with 1383 participants comparing endoscopic resection and laparoscopic resection were identified, while three cohort studies with 167 participants comparing endoscopic resection and laparoscopic and endoscopic cooperative surgery were found. We found that endoscopic resection had shorter operation times (weighted mean difference [WMD] = -27.1 min, 95% confidence interval [CI]: -40.8 min to -13.4 min) and lengths of hospital stay (WMD = -1.43 d, 95% CI: -2.31 d to -0.56 d) than did laparoscopic resection. The results were stable and reliable. There were no significant differences in terms of blood loss, hospitalization costs, incidence of complications or recurrence rates. For tumor sizes 2 - 5 cm, endoscopic resection increased the risk of positive margins (relative risk [RR] = 5.78, 95% CI: 1.31 - 25.46). Although operation times for endoscopic resection were shorter than those of laparoscopic and endoscopic cooperative surgery (WMD = -41.03 min, 95% CI: -59.53 min to -22.54 min), there was a higher incidence of complications (RR = 4.03, 95% CI: 1.57 - 10.34).ConclusionsIn general, endoscopic resection is an alternative method for gastric GISTs ≤ 5 cm. Laparoscopic and endoscopic cooperative surgery may work well in combination. Further randomized controlled trials are recommended to validate or update these results.

Dataset Information

Artificial Intelligence in the Prediction of Gastrointestinal Stromal Tumors on Endoscopic Ultrasonography Images: Development, Validation and Comparison with Endosonographers.

Background/aims

Methods

Results

Conclusions

Publications

Artificial Intelligence in the Prediction of Gastrointestinal Stromal Tumors on Endoscopic Ultrasonography Images: Development, Validation and Comparison with Endosonographers.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets