Chinese scientists have developed an artificial intelligence-powered system for diagnosing upper gastrointestinal cancers through the analysis of endoscopic imaging data.

The system, based on data from more than 50,000 endoscopic images of upper gastrointestinal cancer patients and over 120,000 endoscopic images of non-patients, has registered a 96 percent accuracy rate in diagnosing upper gastrointestinal cancers and 90 percent accuracy in diagnosing such cancers at an early stage.

In the research led by Xu Ruihua, president of the Guangzhou-based Sun Yat-sen University Cancer Center, the system examined more than 1.036 million endoscopic images of 84,424 upper gastrointestinal cancer patients in six hospitals.

An article on the research was published in international science journal Lancet Oncology earlier this month.

This Gastrointestinal Artificial Intelligence Diagnostic System (GRAIDS) was capable of analyzing as many as 118 images per second.

The research team also developed a computer-aided detection (CAD) system for real-time identification of upper gastrointestinal cancerous lesions in routine endoscopic examinations.

The computer with the CAD system was connected to an endoscopy unit, allowing for fully automated diagnostic assistance during endoscopic examinations.

The research team also constructed a cloud-based, multi-institutional AI-fuled platform for patients requiring upper gastrointestinal endoscopies.

It made available a website to provide free access to GRAIDS, allowing clinicians and patients to upload endoscopic images and get a second opinion from GRAIDS.

An open-access endoscopic image database was also made available on the website, which may be useful both for training endoscopists and to researchers in the field of endoscopy and AI-aided medical imaging.

Upper gastrointestinal cancers, including oesophageal cancer and gastric cancer, kill 400,000 patients in China every year.

Most upper gastrointestinal cancers are diagnosed at advanced stages because their signs and symptoms tend to be latent and non-specific, leading to poor prognoses.

In countries with an unbalanced distribution of medical resources between urban and rural areas, GRAIDS can help bridge the cancer diagnosis gap between national hospitals and primary care hospitals, according to the article in Lancet Oncology.