Enhancing Endoscopy Diagnostics with AI Assistance: A Pathway to Better Outcomes
March 07, 2025
Sponsored Story
Doctors are increasingly leveraging AI tools to enhance and advance diagnostics, resulting in improved patient outcomes.
The Challenge: The Struggle to Identify Colorectal Cancer with Accuracy and Human Limitations
Colorectal cancer ranks as the third most common cancer worldwide1. While diagnosing colon-related cancers is traditionally a straightforward procedure, detecting abnormalities along the entire six feet of the large intestine remains a labor-intensive task susceptible to human error.
Research from the National Library of Medicine indicates that physicians fail to spot polyps during colonoscopies between 22% and 28% of the time2. This is not necessarily a failure on the part of the doctors; some adenomas are minuscule, blend into the background, or possess complex morphologies that complicate their identification even in video feeds.
AI has the potential to significantly aid in the process of endoscopy and other diagnostic procedures. Despite the high accuracy of these procedures, doctors may still struggle to achieve the desired levels of accuracy. Technical limitations and human factors, such as fatigue or varying levels of experience among physicians, can also contribute to the inaccuracies observed during endoscopic evaluations.
The Solution: Introducing ASUS EndoAim Endoscopy AI
The ASUS EndoAim Endoscopy AI Solution assists medical practitioners in the real-time identification and classification of polyps in colonoscopy images. The solution tags suspected polyps and other abnormalities with accuracy and sensitivity rates exceeding 90%. It employs a host of AI algorithms and machine vision technology to analyze frames at 60 frames per second with a low latency of less than 70 milliseconds, highlighting the most likely candidates in real-time. While physicians may spot certain anomalies, the AI automatically detects polyps in real-time. This allows doctors to inspect highlighted areas more closely by switching to narrow-band imaging (NBI), where EndoAim can automatically classify the highlighted polyps as adenomas or non-adenomas.
Additionally, EndoAim includes a one-click size measurement feature, replacing traditional visual estimation methods. Leveraging this solution, doctors can increase their polyp detection rates, leading to more accurate evaluations and avoiding unnecessary tissue biopsies or additional procedures.
EndoAim is powered by Intel® Core™ processors, providing high performance for medical AI and serving as a foundation for connected and optimized edge computing systems. It also utilizes the Intel OpenVINO toolkit for improved computer vision models.
The Outcome: The Future of Medical AI Beyond Endoscopy
The results presented by EndoAim have been promising. According to a published paper, CADX/E (AI) can raise the adenoma detection rate about 14%3. Researchers have also noted improvements in the detection rates of small polyps, the time spent on diagnostics, and the overall accuracy of endoscopies. The implementation and system integration of EndoAim with a hospital’s existing facilities are economically friendly, as no additional equipment purchases are required for the upgrade.
System Diagram
ASUS has successfully deployed EndoAim in over 30 medical institutes in Taiwan. These AI-enhanced medical diagnostic systems have potential applications beyond endoscopy. ASUS is working on expanding its computer vision solutions to other aspects of gastrointestinal medicine, such as analyzing imagery from the upper GI tract and the stomach for early cancer detection. The ASUS team is also exploring ways to move beyond diagnostics into predictive medicine, aiding doctors in preventative medicine, which offers significant rewards for patients.
While the advantage of AI in medical diagnostics is already impactful, we are still in the early stages of AI-related medical tools. The potential to expand this technology to ultrasound, X-ray, MRI, CT, and other visual diagnostics is vast. EndoAim is positioned to pioneer this medical transformation.
- Worldwide cancer data | World Cancer Research Fund International (wcrf.org)
- A M Leufkens et al., “Factors influencing the miss rate of polyps in a back-to-back colonoscopy study”, 2012, https://pubmed.ncbi.nlm.nih.gov/22441756/
- Alessandro Repici et al., “Efficacy of Real-Time Computer-Aided Detection of Colorectal Neoplasia in a Randomized Trial”, 2020. https://pubmed.ncbi.nlm.nih.gov/32371116/
