Skip to main content


AI in Healthcare Advances Diagnostic Accuracy and Speed

AI in healthcare

AI in healthcare is changing the face of diagnostic medicine, helping doctors work more accurately to improve patient outcomes.

Use of edge AI in endoscopy procedures is a prime example. Endoscopy involves inserting a tube with a camera (endoscope) into the body to obtain images or video of the patient’s organs and tissues. Endoscopy procedures have multiple uses, but among other things they are a vital diagnostic tool to help gastrointestinal (GI) medicine specialists screen for cancer. Endoscopies allow these physicians to detect polyps, benign but potentially problematic growths, and in particular, adenomas, which are polyps that doctors consider precancerous.

But even the most experienced doctors may be challenged to reliably interpret images from an endoscopy.

“The medical literature tells us that physicians fail to spot polyps during colonoscopies at a rate of 22% to 28%,” says Sabrina Liu, Product Engineer at ASUStek Computer Incorporation, a global developer of diversified computing products. “It’s inherently difficult work: Some adenomas are extremely small and hard to see while polyps have different morphologies that can make them easy to miss on a video feed.”

In addition to the technical challenges of endoscopies, there are also basic human limitations. For example, a doctor at the end of a long shift might be more fatigued and prone to mistakes than at the start of the day. And a junior clinician is unlikely to be as proficient as a more experienced colleague at interpreting medical imagery.

Today’s innovative solutions use edge AI and computer vision to enhance traditional endoscopy equipment. And these systems have already been deployed in real-world clinical settings—with promising results.

Clinical Deployments Demonstrate Improved Accuracy

The ASUS Endoscopy AI Solution EndoAim, currently used at multiple hospitals in Taiwan, is a case in point.

The system highlights AI-detected polyps on the screen in real time by analyzing up to 60 images per second, calling attention to anything the physician may have missed. If they want to inspect a region of interest more closely, they can switch to narrow-band imaging (NBI) and the system will automatically classify selected polyps as adenomas or non-adenomas. Doctors can also use the system to perform one-click measurements of polyps, whereas before they typically determined polyp size by visual judgment, which had a relatively low accuracy of approximately 62.5%.

The results of the solution in clinical settings are impressive. “Physicians have seen their adenoma detection rates improve by 15% to 20% on average,” says Liu. “There is also a significant improvement in detecting small polyps—as well as time savings, because doctors can now measure polyps more quickly and accurately during endoscopies.”

Using #EdgeAI to improve the accuracy and diagnostic consistency of endoscopies will likely appeal to many physicians—and the physical features of these systems add further incentives for adoption. @ASUS via @insightdottech

AI Toolkits, Edge Hardware, and Collaboration Speed Time to Market

Using edge AI to improve the accuracy and diagnostic consistency of endoscopies will likely appeal to many physicians—and the physical features of these systems add further incentives for adoption.

EndoAim is based on a miniature edge PC with a compact form factor of 12cm x 13cm x 5.5cm—a critical consideration in hospital examination rooms where space is at a premium. In addition, the system can be connected to existing endoscopy equipment without specialized medical hardware, making it easier and more cost-effective for clinicians to begin using AI immediately.

The ASUS partnership with Intel was crucial in developing a market-ready product. “Intel CPUs with integrated graphics processing helped us reduce our solution’s overall size—and achieve an image analysis rate of 60 FPS, which is the highest rate currently available to physicians,” says Liu. “Using the Intel® OpenVINO toolkit, we also optimized our computer vision models, enabling them to run more smoothly and efficiently.”

The two companies’ collaboration shows how technology partnerships make it possible to offer powerful solutions to medical device buyers—and do it faster than ever before.

“We started work on EndoAiM in 2019 and had an early model toward the end of 2020, which is when we turned to Intel for engineering support,” says Liu. “By 2021, we had the version of the product that we wanted to take to market.”

The Future of AI in Healthcare: GI Medicine and Beyond

The fact that solutions providers can innovate edge AI systems more quickly and effectively is good news for doctors, patients, and healthcare SIs, as it will no doubt enable other use cases in coming years.

ASUS is already at work on some of those new use cases with its current endoscopy system. Liu says the company plans to expand its computer vision solution to other aspects of gastrointestinal medicine, such as the analysis of imagery from the upper GI tract and the stomach. In addition, ASUS engineers are looking at ways to use AI to build solutions that go beyond detection and diagnostic support and enable the prediction of illness, helping doctors to catch potential problems earlier so patients can begin treatment sooner.

Beyond GI medicine, the underlying computer vision algorithms behind EndoAiM could eventually be applied to other types of medical imaging. “We see the potential to expand this technology to analyzing imagery from ultrasounds, X-rays, MRIs, and more,” says Liu. “There’s a tremendous opportunity to help people here, and we’re excited to hear from clinicians in different medical fields and see how we can develop solutions to meet their needs.”


This article was edited by Georganne Benesch, Editorial Director for