Colorectal cancer (CRC) remains a major global health threat, causing approximately 900,000 deaths annually. Unlike many other cancers, CRC has clear preventive potential: intestinal polyps serve as precancerous lesions that, if undetected, can progress to invasive adenomas or carcinoma. Yet, colonoscopy still faces significant diagnostic challenges. Variations in physician experience or subtle mucosal color differences can lead to missed lesions, while factors such as endoscope rotation and bowel contractions may prevent full mucosal visualization, increasing diagnostic risk.

Addressing this clinical pain point, Dr. Po-Wen Lu’s team at Shuang Ho Hospital, Ministry of Health and Welfare, has developed an “AI-Powered Endoscopic Imaging and Measurement Device” that attaches externally to existing endoscopes. Leveraging artificial intelligence (AI) and deep learning models, the system performs real-time analysis of endoscopic images. The technology can automatically detect lesions, classify their type and location, and provide precise length measurements, offering the potential to improve diagnostic accuracy, reduce physician workload, and advance smart medical solutions toward practical clinical application.

Technical Breakthroughs and Clinical Advantages: Precision Measurement and Real-Time Analysis

Designed as an external add-on, the device interfaces seamlessly with various endoscope models and incorporates edge computing capabilities for immediate image processing. Its core technologies include deep learning-based lesion detection and length measurement algorithms, powered by GPU computing. Performance metrics are notable, with lesion detection achieving an F1-Score of 87.52% and length measurement error as low as 0.87 millimeters. Key advantages include:

• High Accuracy: The deep learning model analyzes texture, brightness, and other features to automatically identify polyps and lesion types, significantly reducing missed and misdiagnoses.
  
• Real-Time Measurement: Physicians can obtain immediate, precise information on lesion length and area, aiding in clinical decision-making.
  
• Edge Computing Efficiency: The external computing platform rapidly processes large image datasets without reliance on cloud computing, ensuring real-time performance and data security.
  
• Market Competitiveness: Designed for high-demand endoscopy centers, the device has completed initial validation at Shuang Ho Hospital in New Taipei City and shows strong potential for international deployment.
  

Looking ahead, the team plans to collaborate with semiconductor developers to convert existing algorithms into a system-on-chip (SOC) solution, supporting pilot and mass production, and expanding into global markets including the U.S., Japan, and Germany — countries with mature endoscopy technology.

Vision for Clinical Application and Market Strategy

Dr. Lu emphasized, “Our goal is to integrate AI-powered analysis into routine endoscopy, combining edge AI technologies developed by Professor Yen-Lin Chen’s team at National Taipei University of Technology to help physicians diagnose faster and more accurately while reducing workload.” Regulatory and risk management planning, covering FDA, CE, and TFDA requirements, has been completed, with comprehensive risk mitigation strategies including algorithm review, data encryption, hardware redundancy, and operator training.

From a commercialization perspective, the team will collaborate with endoscope manufacturers such as Olympus, Fujifilm, and Pentax to integrate AI analysis capabilities and pursue software licensing models. Revenue strategies include single-device sales, software licensing, and regular model updates to ensure sustained product performance and market value.

Future development will focus on further optimizing AI algorithms, expanding clinical applications, and driving international market adoption. By making smart endoscopy accessible, this innovative device not only represents a milestone in medical AI technology but also provides a highly competitive solution for the global smart healthcare market, enhancing early detection and diagnosis of gastrointestinal diseases worldwide.

Resource: 《新創動態》突破大腸鏡檢診斷瓶頸：AI智慧影像裝置應運而生