675

IMPACT OF A REAL-TIME COMPUTER-AIDED POLYP CHARACTERIZATION IN SCREENING COLONOSCOPY PERFORMED BY TRAINEES VERSUS EXPERIENCED ENDOSCOPISTS: A RANDOMIZED CONTROLLED TRIAL

Date
May 20, 2024

Introduction: Although the data on the efficacy of computer-aided optical detection (CADe) in clinical practice is robust, the data on computer-aided optical diagnosis (CADx) for experts and trainees is still lacking. We aimed to fill this gap by evaluating the accuracy, specificity, and negative predictive value (NPV) of CADx during real-time colonoscopy screening compared to the standard colonoscopy performed by experts and trainees.
Methods: This prospective randomized controlled study enrolled asymptomatic subjects undergoing screening colonoscopy, randomly assigned to either the CADx-assisted or standard colonoscopy. The CADx group utilized real-time CADx with a magnified blue light imaging system (CAD-EYE system; Fujifilm Co, Japan), while the standard group underwent high-definition colonoscopy with magnified image-enhanced endoscopy (blue light and narrow band imaging system). The diagnosis made by CADx was recorded as either neoplasia or hyperplastic polyp, while endoscopists in the standard group made the diagnosis. All detected polyps were resected and sent for pathological analysis. Neoplasia was defined as adenoma. The primary outcome was the accuracy of neoplasia detection with a size ≤10 mm. Secondary outcome assessed the influence of endoscopist experience on neoplasia detection accuracy, specificity and NPV.
Results: Between 7/22-8/23, 831 subjects (female 52% and mean age 64±7 years) were enrolled. Of these, 423 were randomized to undergo CADx, and 408 underwent standard colonoscopy. Endoscopists included both 26 attending staff and 21 trainees. Two-thirds of colonoscopies 539 (64.9 %) were performed by trainees. In the CADx group, 572 polyps sized ≤10 mm were resected, with 485 (84%) confirmed as neoplasia. In the standard group, 459 polyps sized ≤10 mm were resected, and 380 (82.7%) were confirmed as neoplasia. The accuracy for neoplasia detection with CADx was 83.7%, compared to 79.5% with the standard group (p = 0.08). Subgroup analysis for trainees, CADx significantly enhanced accuracy to 83.7%, compared to the standard group at 76.8% (p=0.03). Whereas colonoscopies performed by experts revealed comparable accuracy (83.8% vs. 83.2%, p=0.87). CADx increased the specificity significantly (p <0.01, Table 1). NPV in CADx was significantly higher than in the standard group (47.3% vs. 38.1%, p <0.01). CADx demonstrated an improvement of NPV over standard colonoscopy in expert performance (53.7% vs. 42.9%, p = 0.03) but not in trainee performance (41.4% vs. 35.7%, p = 0.15). CADx did not exceed a threshold of 90% NPV and the ESGE requirement of 80% sensitivity and 80% specificity for the resect-and-discard strategy (Table 1).
Conclusion: CADx showed improved accuracy in trainees, matched to experienced endoscopists, but fell short in NPV thresholds for widespread adoption, indicating the need for refinement for a better model.

Tracks

Related Products

Thumbnail for THE IMPROVEMENT ON ADENOMA DETECTION RATE AND OTHER SECONDARY INDICATORS OF THE TWO REAL-TIME ARTIFICIAL INTELLIGENCES IN HIGH ADENOMA DETECTORS: A RANDOMIZED MUTI-CENTER TRIAL
THE IMPROVEMENT ON ADENOMA DETECTION RATE AND OTHER SECONDARY INDICATORS OF THE TWO REAL-TIME ARTIFICIAL INTELLIGENCES IN HIGH ADENOMA DETECTORS: A RANDOMIZED MUTI-CENTER TRIAL
Introduction: Adenoma per colonoscopy (APC) has recently been proposed as a quality measure for colonoscopy. We evaluated the impact of a novel AI system, compared to standard HD colonoscopy, for APC measurement…
Thumbnail for A RANDOMIZED TRIAL OF BALLOON VERSUS NON-BALLOON PERCUTANEOUS ENDOSCOPIC GASTROSTOMY REPLACEMENT TUBE: TUBE LIFESPAN AND MECHANISM OF COMPLICATIONS
A RANDOMIZED TRIAL OF BALLOON VERSUS NON-BALLOON PERCUTANEOUS ENDOSCOPIC GASTROSTOMY REPLACEMENT TUBE: TUBE LIFESPAN AND MECHANISM OF COMPLICATIONS
Endoscopic ultrasound-guided gastroejejunostomy (EUS-GJ) has gained popularity in treating malignant gastric outlet obstruction (GOO). EUS-GJ has also been used to manage benign GOO with promising technical and clinical success…
Thumbnail for THE SUCCESS RATE AND SAFETY OF DOUBLE GUIDEWIRE CANULATION PERFORMED BY TRAINEES AFTER FAILED STANDARD CANULATION
THE SUCCESS RATE AND SAFETY OF DOUBLE GUIDEWIRE CANULATION PERFORMED BY TRAINEES AFTER FAILED STANDARD CANULATION
Training in interventional endoscopy is offered by interventional endoscopy fellowship programs (IEFPs) not accredited by the ACGME. The number of these programs has increased exponentially with a concurrent increase in the breadth and complexity of these procedures…
Thumbnail for A RANDOMIZED TRIAL OF ENDOSCOPIC TRANSPAPILLARY GALLBLADDER STENTING VERSUS EUS-GUIDED TRANSMURAL GALLBLADDER DRAINAGE IN THE PATIENTS WITH ACUTE CALCULOUS CHOLECYSTITIS AND HIGH SURGICAL RISK: THE OUTCOMES AT 18 MONTHS
A RANDOMIZED TRIAL OF ENDOSCOPIC TRANSPAPILLARY GALLBLADDER STENTING VERSUS EUS-GUIDED TRANSMURAL GALLBLADDER DRAINAGE IN THE PATIENTS WITH ACUTE CALCULOUS CHOLECYSTITIS AND HIGH SURGICAL RISK: THE OUTCOMES AT 18 MONTHS
INTRODUCTION: It is unclear whether the insertion of an axis-orienting double-pigtail plastic stent (DPS) through biliary lumen-apposing metal stents (LAMS) in EUS-guided choledochoduodenostomy (CDS) improves the stent patency…