AI-Assisted Detection Identifies Colon Cancer Automatically and in Real-Time

A Much-Needed Improvement

A new computer-aided endoscopic system that probes for signs of tumor or cancer growth in the colon may very well be the future of cancer detection. Assisted by artificial intelligence (AI), the new diagnostic system is able to tell if clumps of cells, called colorectal polyps, that grow along the walls of the colon are benign tumors known as colorectal adenoma.

The computer-assisted diagnostic system was trained using more than 30,000 images of colorectal polyps, each magnified 500-times, and operates using machine learning. The AI can check approximately 300 features of the polyp, which it compares to its existing "knowledge", in less than a second. After having been used successful in preliminary studies, prospective trials followed. The results of these trials, the first for AI-assisted endoscopy in a clinical setting, were presented at the 25th UEG Week in Barcelona, Spain.

The prospective study was conducted by a team led by Dr. Yuichi Mori from Showa University in Yokohama, Japan. Mori and his colleagues tested the new system in 250 patients previously identified to have colorectal polyps. The AI predicted the pathology of each polyp, comparing it with final pathological reports taken from the resected specimens. The results were highly encouraging — the system assessed 306 polyps in real-time, with a 94 percent sensitivity, 79 percent specificity, and 86 percent accuracy. In identifying abnormal tissue growth, the system demonstrated 79 percent positive and 93 percent negative predictive values.

AI in Healthcare

In short, the AI was able to fairly accurately identify which abnormal colon cell growths were most likely to be cancerous. "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," Mori said, speaking during the Opening Plenary at the UEG Week. "This allows the complete resection of adenomatous polyps and prevents unnecessary polypectomy of non-neoplastic polyps."

Furthermore, the researchers presented the results of their prospective study to prove that their system was ready for clinical trials."We believe these results are acceptable for clinical application and our immediate goal is to obtain regulatory approval for the diagnostic system," Mori added.

While this may be the first AI-enabled, real-time biopsy, as Mori described it, it's not the first time AI has been used to improve medical diagnosis and overall medical research. For example, there is an AI effective in identifying skin cancer, and chipmaker NVIDIA is working on a moonshot project to accelerate cancer research using deep learning. Also working in cancer diagnosis is IBM's Watson, which in some cases has proven to be 99 percent accurate in recommending the same treatments as doctors. Improved cancer detection can spell the difference between a treatment that works and one that doesn't, so these advancements are potentially life-saving.

Moving forward, Mori's team plans to conduct a multicenter study to aid eventual clinical tests. They're also working on an automatic polyp detection system. "Precise on-site identification of adenomas during colonoscopy contributes to the complete resection of neoplastic lesions" said Dr Mori. "This is thought to decrease the risk of colorectal cancer and, ultimately, cancer-related death."