Real-time Diagnosis of Diminutive Colorectal Polyps Using AI
- Conditions
- Colorectal NeoplasmsColorectal Polyp
- Interventions
- Device: Computer-aided diagnosis (CADx) systems
- Registration Number
- NCT05349110
- Lead Sponsor
- Maastricht University Medical Center
- Brief Summary
Correct endoscopic prediction of the histopathology and differentiation between benign, pre-malignant, and malignant colorectal polyps (optical diagnosis) remains difficult. Artificial intelligence has great potential in image analysis in gastrointestinal endoscopy. Aim of this study is to investigate the real-time diagnostic performance of AI4CRP for the classification of diminutive colorectal polyps, and to compare it with the real-time diagnostic performance of commercially available CADx systems.
- Detailed Description
Correct endoscopic prediction of the histopathology and differentiation between benign, pre-malignant, and malignant colorectal polyps (optical diagnosis) remains difficult. Despite additional training, even experienced endoscopists continue to fail meeting international thresholds set for safe implementation of treatment strategies based on optical diagnosis.
Multiple machine learning techniques - computer-aided diagnosis (CADx) systems - have been developed for applications in medical imaging within colonoscopy and can improve endoscopic classification of colorectal polyps.
Aim of this study is to explore the feasibility of the workflow using AI4CRP (a CNN based CADx system) real-time in the endoscopy suite, and to investigate the real-time diagnostic performance of AI4CRP for the diagnosis of diminutive (\<5mm) colorectal polyps. Secondary, the real-time performance of commercially available CADx systems will be investigated and compared with AI4CRP performance.
Recruitment & Eligibility
- Status
- UNKNOWN
- Sex
- All
- Target Recruitment
- 105
- Age >18 years;
- Patients with at least one colorectal polyps encountered during colonoscopy;
- Patients referred for a colonoscopy by the Dutch bowel cancer screening program, patients undergoing a colonoscopy for endoscopic surveillance, or patients undergoing a colonoscopy because of complaints;
- Written informed consent.
- Patients with prior history of inflammatory bowel diseases (IBD) or polyposis syndromes;
- Patients with inadequate bowel preparations after adequate washing, suctioning, and cleaning manoeuvres have been performed by the endoscopist;
- Patients undergoing an emergency colonoscopy;
- Written objection in the patient file for participation in scientific research.
Study & Design
- Study Type
- OBSERVATIONAL
- Study Design
- Not specified
- Arm && Interventions
Group Intervention Description Gastroenterology patients Computer-aided diagnosis (CADx) systems Patient receiving a colonoscopy because of regular care will be considered eligible for inclusion if at least one diminutive colorectal polyp is encountered during the colonoscopy. Patients receive an endoscopic procedure in the context of the Dutch national screening program, because of gastrointestinal symptoms, or because of follow-up of previously diagnosed bowel diseases. Colonoscopies will be executed using Fujifilm endoscopy systems (Fujifilm® Corporation, Tokyo, Japan), using Pentax endoscopy systems (Pentax Medical®, Hamburg, Germany), and using Olympus endoscopy systems (Olympus®, Tokyo, Japan).
- Primary Outcome Measures
Name Time Method The negative predictive value of AI4CRP per image modality (HDWL, BLI, LCI, i-scan). 1 year The real-time negative predictive value of AI4CRP per image modality (HDWL, BLI, LCI, i-scan).
Technical feasibility of real-time use of AI4CRP. 6 months The technical feasibility of real-time use of AI4CRP in the endoscopy suite regarding a proper reception of the video output from the local endoscopy processor towards AI4CRP (in high definition quality, without any delays in time).
The specificity of AI4CRP per image modality (HDWL, BLI, LCI, i-scan). 1 year The real-time specificity of AI4CRP per image modality (HDWL, BLI, LCI, i-scan).
User interface feasibility of real-time use of AI4CRP. 6 months The user interface feasibility of real-time use of AI4CRP in the endoscopy suite regarding a correct alignment of the user interface of AI4CRP with the video output from the local endoscopy system (resizing image pixels and anonymization).
The diagnostic accuracy of AI4CRP per image modality (HDWL, BLI, LCI, i-scan). 1 year The real-time diagnostic accuracy of AI4CRP per image modality (HDWL, BLI, LCI, i-scan). Diagnostic accuracy defined as the percentage of correctly optically diagnosed colorectal polyps.
The sensitivity of AI4CRP per image modality (HDWL, BLI, LCI, i-scan). 1 year The real-time sensitivity of AI4CRP per image modality (HDWL, BLI, LCI, i-scan).
The Area Under ROC Curve (AUC) of AI4CRP per image modality (HDWL, BLI, LCI, i-scan). 1 year The real-time Area Under ROC Curve (AUC) of AI4CRP per image modality (HDWL, BLI, LCI, i-scan).
The positive predictive value of AI4CRP per image modality (HDWL, BLI, LCI, i-scan). 1 year The real-time positive predictive value of AI4CRP per image modality (HDWL, BLI, LCI, i-scan).
- Secondary Outcome Measures
Name Time Method The sensitivity of AI4CRP per polyp. 1 year The real-time sensitivity of AI4CRP per polyp (comprising the combination of different imaging modalities).
The sensitivity of CAD EYE in BLI mode, per polyp. 1 year The real-time sensitivity of CAD EYE in BLI mode, per polyp.
The localization score of AI4CRP. 1 year The localization score of AI4CRP regarding the number of images in which the heatmap produced by AI4CRP pointed out the area of interest (scale: correct, incorrect, or partly correct area of interest).
The diagnostic accuracy of CAD EYE in BLI mode, per polyp. 1 year The real-time diagnostic accuracy of CAD EYE in BLI mode, per polyp.
The negative predictive value of CAD EYE in BLI mode, per polyp. 1 year The real-time negative predictive value of CAD EYE in BLI mode, per polyp.
The specificity of AI4CRP per polyp. 1 year The real-time specificity of AI4CRP per polyp (comprising the combination of different imaging modalities).
The Area Under ROC Curve (AUC) of CAD EYE in BLI mode, per polyp. 1 year The real-time Area Under ROC Curve (AUC) of CAD EYE in BLI mode, per polyp.
The difference in diagnostic accuracy of endoscopists per polyp before and after AI. 1 year The difference in real-time diagnostic accuracy of endoscopists per polyp before and after AI.
The difference in sensitivity of endoscopists per polyp before and after AI. 1 year The difference in real-time sensitivity of endoscopists per polyp before and after AI.
The diagnostic accuracy of AI4CRP per polyp. 1 year The real-time diagnostic accuracy of AI4CRP per polyp (comprising the combination of different imaging modalities).
The positive predictive value of AI4CRP per polyp. 1 year The real-time positive predictive value of AI4CRP per polyp (comprising the combination of different imaging modalities).
The Area Under ROC Curve (AUC) of AI4CRP per polyp. 1 year The real-time Area Under ROC Curve (AUC) of AI4CRP per polyp (comprising the combination of different imaging modalities).
The diagnostic accuracy of AI4CRP per patient. 1 year The real-time diagnostic accuracy of AI4CRP per patient (in case of multiple polyps per patient).
The diagnostic accuracy of CAD EYE per patient. 1 year The real-time diagnostic accuracy of CAD EYE per patient (in case of multiple polyps per patient).
The difference in specificity of endoscopists per polyp before and after AI. 1 year The difference in real-time specificity of endoscopists per polyp before and after AI.
The negative predictive value of AI4CRP per polyp. 1 year The real-time negative predictive value of AI4CRP per polyp (comprising the combination of different imaging modalities).
The specificity of CAD EYE in BLI mode, per polyp. 1 year The real-time specificity of CAD EYE in BLI mode, per polyp.
The positive predictive value of CAD EYE in BLI mode, per polyp. 1 year The real-time positive predictive value of CAD EYE in BLI mode, per polyp.
The difference in positive predictive value of endoscopists per polyp before and after AI. 1 year The difference in real-time positive predictive value of endoscopists per polyp before and after AI.
The agreement in surveillance interval based on optical diagnosis and histopathology. 1 year The agreement in surveillance interval based on optical diagnosis of diminutive colorectal polyps and histopathology of small and large colorectal polyps, compared to the surveillance interval based on histopathology of all colorectal polyps (diminutive, small, and large).
The difference in negative predictive value of endoscopists per polyp before and after AI. 1 year The difference in real-time negative predictive value of endoscopists per polyp before and after AI.
Trial Locations
- Locations (2)
Maastricht University Medical Center
🇳🇱Maastricht, Limburg, Netherlands
Catharina Ziekenhuis Eindhoven
🇳🇱Eindhoven, Noord-Brabant, Netherlands