▲ Multi-center clinical results involving 35 endoscopists published in international journal

▲ Colonoscopy diagnostic accuracy improves from 78.3% to 89.8% with AI assistance

▲ Greatest effect seen in novice endoscopists — suggests potential for 'standardizing' endoscopy quality

Ainex Corporation (CEO David Lee), a medical AI company, has released findings from a clinical study using its AI-based real-time colonoscopy diagnostic assistance software, ENAD CADx.

The study, a multi-center trial examining the impact of AI diagnostic assistance on endoscopists' diagnostic accuracy, has been published in the international gastroenterology journal Gut and Liver. The key finding is that ENAD CADx significantly improves endoscopist performance, with the greatest benefit observed in the novice group. The study is notable for demonstrating how much AI can contribute to reducing variation in optical diagnosis quality driven by differences in endoscopist experience.

Optical diagnosis involves real-time observation of lesion morphology during endoscopy to estimate histological type. It has attracted attention for its potential to reduce the risks and costs associated with unnecessary polyp resection.

However, for optical diagnosis to be effectively applied in clinical practice, it must meet international benchmarks such as SODA (Simple Optical Diagnosis Accuracy) and PIVI (Preservation and Incorporation of Valuable Endoscopic Innovations). These standards demand high diagnostic accuracy and negative predictive value (NPV), making them difficult for novice endoscopists to achieve.

The study enrolled 35 endoscopists from four institutions: Seoul National University Hospital Gangnam Center, Seoul National University Hospital, Pusan National University Hospital, and Inje University Busan Paik Hospital. Participants were classified into three groups — trained experts, experts, and novices — based on their optical diagnosis training and experience. Each endoscopist evaluated 100 colorectal lesions (51 adenomas, 39 hyperplastic polyps, 10 sessile serrated lesions) presented as narrow-band imaging (NBI) video clips, with and without ENAD CADx assistance.

Results showed that ENAD CADx assistance enabled faster high-confidence judgments based on real-time AI analysis, leading to improved diagnostic accuracy. The research team defined a "high-confidence (HC) diagnosis" as a judgment made without hesitation within 3 seconds of observing a lesion. With ENAD CADx, HC diagnostic accuracy rose from 78.3% to 89.8% — an increase of 11.5 percentage points. Average decision-making time also decreased from 3.3 to 2.8 seconds, reflecting improved diagnostic efficiency.

The effect was most pronounced in the novice group. Novices' HC diagnostic accuracy increased from 74.1% to 88.8% — a gain of approximately 14.7 percentage points — and inter-rater diagnostic agreement improved significantly (κ = 0.37 → 0.65). This suggests that ENAD CADx can function as a "training equalizer," reducing experience-driven variation in diagnostic quality.

ENAD CADx also simultaneously improved both the accuracy and rate of high-confidence diagnoses, indicating its potential to help endoscopists meet benchmarks such as PIVI and SODA — a factor seen as enhancing the feasibility of optical diagnosis in real-world clinical settings.

Ainex CEO David Lee commented: "This multi-center study provides important evidence that ENAD CADx goes beyond simply improving diagnostic accuracy — it contributes to raising the overall standard of endoscopy quality. We will continue to build a body of clinical evidence to establish ENAD CADx as a trusted endoscopy AI solution in the global market and to set a new standard for colonoscopy diagnosis."

ENAD CADx is the world's first AI colonoscopy diagnostic assistance software capable of real-time three-class classification of lesions into adenoma/neoplasm, hyperplastic polyp, and sessile serrated lesion. Last September, it became the first colonoscopy AI to pass Korea's Innovative Medical Device integrated review, and starting January 2026, non-reimbursed billing as an innovative medical technology became available at general and tertiary hospitals — rapidly expanding its reach across clinical practice.