AI can increase success rates for cancer detection, says report

Japanese researchers train AI to spot cancer early using 30,000 photos of cancerous and per-cancerous tumours

A group of researchers based in Japan have used artificial intelligence to successfully detect colorectal cancer before the tumours become malignant.

The study was led by Dr Yuichi Mori at Showa University, and the findings were presented at the United European Gastroenterology conference. It took place in Barcelona on Sunday.

Using an AI powered tool, the researchers focused on a colorectal polyp that was magnified by 500 times. The tool was able to spot its variations and any other interesting details.

When this stage was completed, the tool compared these variations to results from a database of more than 30,000 photos of precancerous and cancerous cells.

Experts used the photos to train the technology. With this information, the machine learning programme was able to come to a diagnosis within less than a second.

Throughout the study, there was an overall accuracy rate of 86 per cent, and this is impressive considering that the patients who were observed had already been diagnosed.

The AI programme investigated 306 polyps in total. In other findings, it provided sensitivity of 94 per cent, a specificity of 79 per cent, and positive and negative predictive values of 79 per cent and 93 per cent.

This is seen as a crucial study because colorectal cancer is one of the most deadliest forms of cancer, behind lung cancer. It's particularly deadly because, in the later stages, infected cells can travel to other parts of the body through the bloodstream.

Dr Mori's groundbreaking study could see an increase in survival rates for patients with this form of cancer, and he wants to start deploying this technology soon.

Study leader Dr Yuichi Mori said: "The most remarkable breakthrough with this system is that artificial intelligence enables real-time optical biopsy of colorectal polyps during colonoscopy, regardless of the endoscopists' skill.

"This allows the complete resection of adenomatous (cancerous) polyps and prevents unnecessary polypectomy (removal) of non-neoplastic polyps.

"We believe these results are acceptable for clinical application and our immediate goal is to obtain regulatory approval for the diagnostic system."

Speaking to the Telegraph, Cancer Research UK's Dr Claire Knight said artificial intelligence and other emerging technologies offer health professionals a lot of potential.

"AI and virtual reality are opening up many exciting areas of exploration to increase our understanding and treatment of cancer," she said.

"The technology in this presentation could help reduce the overtreatment of bowel growths, called polyps, by helping doctors decide if they need removing or can be left alone. But it will need testing in much larger groups of people first before we understand it's potential."