What Is AI-Assisted Endoscopy Diagnosis? A Guide to GI Endoscopy AI Software

AI-assisted endoscopy diagnosis, also called computer-aided detection and diagnosis (CAD) for endoscopy, is one of the fastest-growing applications of medical imaging AI. In gastroscopy and colonoscopy, an AI software analyzes the endoscopic video stream in real time, highlights suspicious regions, and helps physicians produce complete, standardized reports.

yoyolink AI GI Endoscopy System (hardware + software, 内窥镜影像辅助诊断系统) is such a system, co-developed with Class-A tertiary hospitals in China, holding an NMPA medical device registration certificate and validated through multicenter clinical studies with 1000+ cases.

How AI-assisted GI endoscopy works

During an examination, the software connects to the endoscopy device through standardized interfaces and receives the live HD video signal (up to 1080p@60fps with H.264/H.265 hardware encoding). Traditional deep learning models then process every frame in real time to:

• Automatically detect whether the current examination is a gastroscopy or a colonoscopy, and switch modes accordingly;
• Identify and annotate over 80 types of imaging features of digestive tract anatomical sites and common lesions — including early-stage cancer, polyps and ulcers;
• Track anatomical sites such as the duodenal bulb, gastric fundus and colon, so physicians can confirm coverage and avoid missing key areas;
• Auto-capture screenshots of suspected lesions for the report and follow-up review;
• Record insertion and withdrawal times and score examination quality, including the Boston Bowel Preparation scale for colonoscopy.

From images to a structured image-text report

One of the biggest time costs in an endoscopy center is report writing. yoyolink automatically selects 8–12 key images per examination and generates a structured image-text report with complete, objective descriptions that follow standard requirements. The physician reviews and confirms the report before it is issued — typically a 200KB image-text report is generated in under one second.

The reports and image-text information upload to the hospital storage server through HIS/PACS integration, forming a closed-loop information security model with role-based access control and audit logs.

Why "no AI hallucination" matters

Many consumer AI products generate content with large language models (LLMs), which can "hallucinate" — producing fluent but incorrect output. In a clinical context, that is unacceptable. yoyolink is built on self-developed deep learning models trained on hundreds of thousands of real clinical endoscopic images: every recognition result and description comes from real imaging data, so the output is objective, reliable and traceable. Physicians always remain the final decision-maker; the software is an assistive reference, not a standalone basis for diagnosis.

Learn more

• Product overview — features, modules and technical architecture
• Clinical applications — early cancer screening, QC statistics and deployment scenarios
• Contact Basemed Medical — book a demo or request materials

How patented processing supports reporting

Image recognition is only half of endoscopy AI — the other half is how reports are generated and routed. The invention patent "Information Processing Method and Device for Endoscopic Examination Report" (CN112233752A/B, co-filed with Zhejiang University) covers this: structured lesion recognition, criticality assessment, report print priority, and neural-network matching of patients to suitable physicians. A second patent, "Method and System for Intelligent Transmission of Endoscopic Images Based on Block Decoding" (CN112714233B), secures image transfer with block-based management, multi-layer keys and transmission quality detection — supporting HIS/PACS integration and hospital-local storage.

Patents & IP

Basemed Medical protects its technology with granted invention patents:

• "Information Processing Method and Device for Endoscopic Examination Report" (CN112233752A/B, co-filed with Zhejiang University): intelligently processes endoscopy reports — detects lesions, assesses criticality, prioritizes report printing, and matches patients with suitable physicians via neural networks. View | PDF
• "Method and System for Intelligent Transmission of Endoscopic Images Based on Block Decoding" (CN112714233B): transmits endoscopic images in blocks with multi-layer key information and quality detection, improving transmission quality, security and efficiency. View | PDF

See the full portfolio: About · Patents & IP