Intel and Taiwanese computer company Asus are using artificial intelligence (AI) to help doctors detect cancer.

The AI-boosted endoscopy powered by Intel can help doctors identify colon polyps faster and more accurately.

Colorectal cancer is the third most-common cancer worldwide, with almost two million new cases detected in 2020, according to World Cancer Research.

While diagnosing colon-related cancers is a relatively straightforward procedure – most adults will undertake a colonoscopy by the time they’re 60 – spotting abnormalities along all two metres of the large intestine is still a labour-intensive process. Backed by years of experience, doctors performing the procedure rely on a steady hand, sharp eyesight and unwavering attention to detail.

But you’re still dealing with humans, and that means errors. Detection rates for adenomas, a type of growth along the colon wall, tend to decrease through the course of the day. That’s largely due to human-related causes, according to a 2018 study, as operators rush procedures or because of plain human fatigue.

And the risk of a miss is alarming in even a small percentage of cases when you consider as many as 15 million colonoscopies are performed in the USA each year, according to CNN.

However, a collaboration between Asus and Intel has led to an AI-boosted method that spots adenomas in mere milliseconds, often before the human eye can.

Launched in November, the Asus EndoAim AI endoscopy system is about as large as the palm of a hand. It sits unobtrusively in a colonoscopy room, blending into the background and clutter of other, larger medical equipment. It feeds video data from the colonoscopy camera through an Asus mini PC running on an 11th-generation Intel Core processor and outputs the result at a smooth 60 frames per second.

The device tags suspected polyps and other abnormalities on the screen by bracketing them in green, and immediately classifies the polyps, saving medical staff the guesswork. And it doesn’t just tag a single polyp at a time. Asus claims that in clinical trials, EndoAim can bracket up to 50 polyps in real time.

Results aren’t confined to clinical trials.

Late last year, a Taiwan doctor using an early version of EndoAim, said it helped him detect eight polyps simultaneously during a routine colonoscopy. In early March 2023, another doctor thanked EndoAim for helping him locate a 0.3cm adenoma hidden in the deepest part of a patient’s large intestine.

Today, EndoAim is used at five medical facilities in Taiwan – the Kaohsiung Veterans General Hospital, National Yang Ming Chiao Tung University Hospital, Dianthus Medical Group, Dr Lee Clinic and Ansn Clinic. 

Its success isn’t due just to smart silicon in a compact case. Intel engineers helped Asus fine-tune their AI to detect the tiniest polyp on a wide array of Intel open-source software, including the OpenVino toolkit. The ability to write code seamlessly and quickly to accelerate the system’s development and bring it to market in less than two years “played a crucial role in bringing our vision to life”, said April Yang, project manager for EndoAim at Asus.

“Our goal is to showcase the effectiveness of EndoAim in Taiwan and emphasise how AI can truly assist physicians,” said Yang. “Just like advanced driver-assistance systems in cars, AI does not replace the physicians but serves as a reliable backup, enabling high-quality inspections,”

She added that EndoAim differentiated itself from other AI-boosted offerings by offering real-time polyp detection with up to 95% sensitivity and the ability to classify polyps with up to 95% AUC, a type of performance measurement in machine-learning models.

Other offerings require expensive add-on scopes to enable real-time polyp detection, said Yang. EndoAim doesn’t require any additional hardware beyond plugging the mini PC into the scope’s camera.

“Ultimately, we aim to help doctors increase the polyp detection rate during scope-related procedures and to help them make accurate evaluations that avoid unnecessary tissue biopsies,” she said.

When Asus engineers began developing this in 2021 with AI computing technology integrated into a large tower-based PC chassis, they quickly realised space was a huge constraint.

Intel suggested the 11th-generation CPU family. The Intel CPUs feature Intel Iris X^e Graphics technology that provides discrete-level integrated graphics.

“With OpenVino, Asus found they were able to achieve 60 frames per second detection rate within a compact Intel CPU,” said Tasha Chuang, director of the Asus account team at Intel in Taiwan. “That also means lower cost, and they were able to enable a fanless design as well as a smaller footprint, which makes the portability possible.”

Asus and Intel are exploring opportunities to enhance EndoAim by boosting its display output and computing prowess, and to expand the technology into other medical areas.

“We need more computing power and are currently testing Intel’s 13th-generation processor family now,” said Yang, explaining how Asus was looking to evolve future options with higher 4K resolutions.

“We’re also exploring the ability to add more functionality that doctors and medical staff will find useful, and we look forward to collaborating with Intel in this area,” she said. “This is just the beginning.”