Evaluation of CAD-EYE/SCALE-EYE for Detection, Classification, and Measurement of Colorectal Polyps: a Prospective Study

Evaluation of CAD-EYE/SCALE-EYE for Detection, Classification, and Measurement of Colorectal Polyps: a Prospective Study

Performance Evaluation of CAD-EYE and SCALE-EYE for Detection, Classification, and Measurement of Colorectal Polyps: a Prospective Study

The investigators hypothesize that the clinical implementation of an AI system is an optimal tool to monitor, audit and improve the detection and classification of polyps during colonoscopy. The purpose of this prospective clinical cohort study is to evaluate the performance of the SCALE-EYE virtual scale for measuring polyp size when used during live colonoscopies. The investigators also wish to evaluate CAD-eye for detection and classification of polyp histology. It is hypothesized that CAD-eye and SCALE-EYE can function in real-time practice with high accuracy.

Detailed Description: This study will be conducted in several phases: phase 1 will evaluate the feasibility of using CAD-eye in real-time practice, training endoscopists to work with this platform, obtaining preliminary data on relative size measurement accuracy when using scale eye, testing feasibility of fresh polyp specimen size measurement and determining the sample size and reference standard for evaluating SCALE-EYE during the subsequent phases of the study in randomized controlled trials. Phase 2 will evaluate relative size measurement accuracy of SCALE-EYE compared to visual size estimation and CADe/CADx performance in a randomized controlled trial. Scale-eye will further be validated through video-based analysis comparing different size measurement methods (scale eye, snare, forceps, visual assessment) with different groups of endoscopists (experts, staff gastroenterologists, trainees evaluating videos showing video sequences that show the polyps either with scale eye, visual or instrument information. Furthermore, fresh specimen size measurement and pathology-based size (including different ways to prepare specimens for pathology sectioning) will be evaluated for agreement between size measurement methods and tissue shrinkage effects. Phase 3 will evaluate relative size measurement accuracy of SCALE-EYE compared to snare-based size estimation. Eligible patients will be adults (aged 45 to 80 years) referred to CHUM for diagnostic, screening or surveillance colonoscopy. Patients will undergo colonoscopies using CAD-eye, and the size of all polyps detected will be estimated by different subjective and objective methods. All polyp size measurements will be videorecorded. Video recordings with different measurement instruments (forceps, snare, scale eye, visual assessment) will be assessed with different groups of endoscopists (experts, staff gastroenterologists, trainees) for size measurements through online video presentations in a randomized order. The main results of this study are to evaluate the relative accuracy of SCALE-EYE for polyp size estimation and the adenoma detection rates, sensitivity, specificity, NPV for CAD-eye for polyp detection and classification. We will also calculate other characteristics of CAD-eye and SCALE-EYE tests, the rate of divergence and/or agreement between subjective and objective methods of polyp size estimation, the agreement between surveillance intervals based on optical diagnosis by CAD-eye and pathological findings, and the false positive rate and latency for polyp detection.

The randomization would be concealed. The patient will be blind to the allocated groups. The endoscopist will be blind to the allocated groups until the beginning of the procedure. The pathologists will be blinded to the polyp size measurements by the endoscopists, the SCALE-EYE, or other methods.

Colonoscopies will be performed according to the standard of care. Patients will undergo colonoscopies using CAD-eye (CADe and CADx) and Scale-eye. All optically diagnosed (CADx) polyps will be removed and sent to the CHUM pathology laboratory for histopathological evaluation according to institutional standards. CADx will be combined with human input for final diagnosis.

Real time polyp detection using a standard colonoscopy without CAD eye and without CADe. Patients will undergo a standard colonoscopy. Polyp size measurement will be assessed visually (phase 1 and 2) or with instruments such as snare or forceps (phase 3 or later) and polyp classification will be done by CADx without further input by the endoscopist. All optically diagnosed polyps will be removed and sent to the CHUM pathology laboratory for histopathological evaluation according to institutional standards.

Detection and classification and size measurement of polyp by Artificial Intelligence (Cad eye) and the Scale Eye

Detection and classification and size measurement of polyp by Artificial Intelligence (CADeye) and Scale Eye This study will be conducted in multiple phases: phase 1 (pilot phase, n=40 polyps, about 60 patients) to evaluate the feasibility of applying Scale Eye in real-time practice, training endoscopists to work with CADe and CADx and determining the sample size and reference standard to evaluate SCALE-EYE during the second phase of the study (randomized controlled trial), phase 2 will evaluate the performance of SCALE-EYE and CAD-eye in real-time practice. The sample size for phase 2 (RCT comparing visual with scasle eye) will be based on the pilot data allowing to compare relative accuracy of size measurement with scale eye versus visual size estimation. After completion of phase 2 later phases will use instruments such as snare and/or forceps as reference measurement devices compared to relative accuracy of measurement when using scale eye.

to compare the relative accuracies of SCALE-EYE size measurement, visual size estimation or size estimation based on a standard polypectomy snare for colorectal polyps during real-time colonoscopies with measurement of viable specimens by a Vernier caliper as reference standard. Relative accuracy is defined as "1-[(ScaleEye measurement - gold standard measurement)/gold standard measurement] x100". Scale: 0-100. Higher score is a better outcome.

compare the adenoma detection rate between the CAD-eye-assisted colonoscopy and standard colonoscopy (i.e., detecting adenomas only by endoscopists)

evaluate the accuracy, sensitivity, specificity, positive and negative predictive value of CADx for classifying 1-5 mm polyp histology into neoplastic and non-neoplastic compared with the histopathology outcomes as the reference standard

compare in video-based assessments with different endoscopists (expert level, staff GI, surgeons, trainees the relative size measurements accuracy of scale eye with visual and instrument based (snare and/or forceps) size measurement accuracy

evaluate the sensitivity of the CAD-eye in combination with adjunct methods for real-time detecting colorectal polyps compared with standard colonoscopy

compare the accuracy, sensitivity, specificity, positive and negative predictive value of CADx for classifying 1-5 mm polyp histology to that of using human decision based on CADx suggestion

calculate the proportion of 1-5 mm polyps, CADx can not suggest histology (in the intervention arm, please refer to the randomization section), while endoscopists can suggest histology with high confidence, or vice versa

evaluate diagnostic uncertainty of the current CADx module when predicting the pathology of colorectal polyps

evaluation of the relative accuracy of 3D scanned polyps compared to measurement of viable specimens by a Vernier caliper as reference standard. Relative accuracy is defined as "1-[(ScaleEye measurement - gold standard measurement)/gold standard measurement] x100". Scale: 0-100. Higher score is a better outcome.

evaluation of the relative accuracy of histologic size measurement compared to measurement of viable specimens by a Vernier caliper as reference standard. Relative accuracy is defined as "1-[(ScaleEye measurement - gold standard measurement)/gold standard measurement] x100". Scale: 0-100. Higher score is a better outcome.

Inclusion Criteria: indication of undergoing a screening, surveillance, or diagnostic colonoscopy Age 45-80 years Signed informed consent Exclusion Criteria: Known inflammatory bowel disease (IBD) Active Colitis Poor general health, defined as an American Society of Anesthesiologists (ASA) physical status class >3 Coagulopathy Familial polyposis syndrome Emergency colonoscopies Poor bowel preparation score defined as the total Boston bowel preparation score <6 and the right-segment score <2. For per-polyp analysis: For optical diagnosis (CADx) we will exclude all polyps >5mm. We will also exclude sessile serrated polyps and traditional serrated adenomas from the analysis as CADx can only differentiate between adenomas and hyperplastic polyps. For polyp size measurement (SCALE-EYE), we will exclude all polyps resected in several pieces (piecemeal resection), or specimens damaged (e.g., fractured) during the suction and retrieval process, submillimeter polyps too small for measurement after resection. Polyps after the 5th resected polyp in a single piece will be excluded to reduce the time associated with the procedure. For polyp detection (CADe) we will not exclude any polyp from the analysis.