ARTEMIS: Artificial-intelligence-based Reporting Technology for Endoscopy Monitoring and Imaging System
Study Details
Study Description
Brief Summary
Properly documenting withdrawal time in colonoscopy is essential for quality assessment and cost allocation. However, reporting withdrawal time has significant interobserver variability. Additionally, current manual documentation of endoscopic findings is time-consuming and distracting for the physician. This trial examines an artificial intelligence based system to determine withdrawal time and create a structured report, including high-quality images (AI) of detected polyps and landmarks.
| Condition or Disease | Intervention/Treatment | Phase |
|---|---|---|
|
Detailed Description
This study aims to compare withdrawal time precision calculated by an AI system with examiner-reported times during colonoscopy, also evaluating endoscopists' satisfaction with the images included in the AI-generated reports. The study will be single-center and endoscopist-blinded, where 138 patients are expected to be recruited, taking polyp detection rates and potential dropouts into consideration. Manual annotation of withdrawal times from examination recordings will establish gold standard annotations. The AI system performs a frame-by-frame analysis of endoscopy recordings, predicting endoscopic findings. Using a rule-based logic, the method calculates withdrawal time for the examination and automatically generates a report for the examination. The study will include consenting adult patients eligible for colonoscopy, excluding those meeting specific criteria.
In this observational study, the withdrawal time for the examinations of all recruited patients is estimated by both the physician and the AI method. The study does not relate to any particular indication, and any patient that is appointed for a colonoscopy and does not meet the exclusion criteria can be recruited. The AI method operates in the background, having no influence on the examination's process, or outcome. The standard procedure requires physicians to estimate the withdrawal time and document it in the examination report. Simultaneously, the proposed AI method also computes the withdrawal time for all patients in the background, without affecting the physician, the examination, or the outcomes of the examination. Importantly, the physician remains blinded to the AI model's output.
To establish the gold standard withdrawal time, manual calculations will be performed using the recorded examination data for all patients. This gold standard is used for evaluating errors in withdrawal time estimation made by both the physician and the AI method. Subsequently, a comparative analysis is conducted to assess the disparities between the physician's estimations and those of the AI method.
Furthermore, the AI method captures characteristic images of anatomical landmarks and notable events, such as polyp resections, during the examination. A panel of certified endoscopists will rigorously evaluate the quality and relevance of these selected images.
Study Design
Arms and Interventions
| Arm | Intervention/Treatment |
|---|---|
| Experimental: Intervention arm All patients within the study are included in the intervention arm: The withdrawal time for the interventions for all patients is documented by the physician and the proposed AI system. |
Device: EndoMind
Withdrawal time is calculated and an image report is generated using the EndoMind system.
|
Outcome Measures
Primary Outcome Measures
- Withdrawal time error comparison for colonoscopies using the proposed AI system versus physician estimation [Through study completion, an average of 5 months]
The error between gold standard withdrawal time and the withdrawal time estimated from the proposed AI system and the physician are compared for the same examination.
Secondary Outcome Measures
- Image quality satisfaction [Through study completion, an average of 5 months]
A board of endoscopy experts receives the reports generated by the proposed system and evaluate the quality and satisfaction for the images included in the report. Evaluation will be performed on a Likert scale from 1 to 5.
- Subgroup analysis for withdrawal time calculation error based on the presence or absence of resections in the examination. [Through study completion, an average of 5 months]
Interventions will be split into two categories, those containing at least one resection and those without any resections.The error between gold standard withdrawal time and the withdrawal time estimated from the proposed AI system and the physician are compared for each subgroups of examinations separately.
- Number of examination where withdrawal time could not be determined [Through study completion, an average of 5 months]
Evaluation of the number of examinations where the withdrawal time could not be calculated and the causes.
Eligibility Criteria
Criteria
Inclusion Criteria:
-
Adult patients (>18 years)
-
Scheduled for colonoscopy
Exclusion Criteria:
Patient / Examination level
-
Inflammatory Bowel Disease
-
Familial Polyposis Syndrome
-
Patient after radiation/resection of colonic parts
Data level
-
Endoscopic recordings started after beginning of withdrawal.
-
Examination recordings stopped before the end of the examination.
-
Examinations with corrupt video signal
Contacts and Locations
Locations
| Site | City | State | Country | Postal Code | |
|---|---|---|---|---|---|
| 1 | Universitätsklinikum Würzburg | Würzburg | Bayern | Germany | 97080 |
Sponsors and Collaborators
- Wuerzburg University Hospital
Investigators
None specified.Study Documents (Full-Text)
None provided.More Information
Publications
None provided.- AI03