Intestinal metaplasia in the stomach C16 K29.4 B98.0
Conditions
Interventions
Group 1: Included are adult patients with gastric carcinoma or precancerous gastric lesions for whom endoscopic image and/or video data from routine clinical diagnostics as well as a corresponding his
Sponsors
Universitätsklinikum Augsburg
Eligibility
Sex/Gender
All
Age
18 Years to No maximum
Inclusion criteria
Inclusion criteria: 1. Age = 18 years 2. Gastric cancer or precancerous gastric lesion 3. Availability of endoscopic image and/or video data 4. Availability of a corresponding histopathological reference standard 5. Sufficient image/video quality for algorithmic analysis 6. In gastric cancer cases: histology from endoscopic resection or surgical specimen 7. In surgically treated cases: endoscopic imaging obtained no more than 30 days before surgery 8. Data derived from the period
Exclusion criteria
Exclusion criteria: 1. Age < 18 years 2. Missing endoscopic image and/or video data 3. Missing corresponding histopathological reference standard 4. Insufficient image or video quality for algorithmic analysis 5. In gastric cancer cases: biopsy-only confirmed cases without resection or surgical histology 6. Surgically treated gastric cancer cases if the corresponding endoscopic imaging was obtained more than 30 days before surgery 7. Missing or insufficient linkage between endoscopic data and histopathological findings 8. Presence of direct personal identifiers in the datasets to be analyzed if anonymization or pseudonymization for study use is not possible
Design outcomes
Primary
| Measure | Time frame |
|---|---|
| 1. Outcome: Diagnostic accuracy of the developed AI algorithm for the detection and characterization of gastric cancer and precancerous gastric lesions in endoscopic image and video data compared with the histopathological reference standard (sensitivity, specificity, AUROC). 2. Timepoint of assessment: After completion of algorithm training, during the validation and testing phase within the study period. 3. Method of assessment: Computer-based analysis of pseudonymized endoscopic image and video data and corresponding histopathological reference findings derived from routine clinical care and prospective study data. | — |
Secondary
| Measure | Time frame |
|---|---|
| 1. Diagnostic accuracy of the AI algorithm for endoscopic quality control (sensitivity, specificity, AUROC, accuracy). 2. Performance of automated lesion segmentation in endoscopic image and video data (e.g., Dice coefficient and related segmentation metrics). 3. Comparative diagnostic performance of the AI algorithm versus experienced endoscopists (sensitivity, specificity, AUROC, agreement measures where applicable). 4. Retrospective quality assessment of endoscopic examinations based on predefined AI-derived quality indicators. | — |
Countries
Germany
Contacts
Public ContactHelmut Messmann
Universitätsklinikum Augsburg
Outcome results
None listed