AI vs Cancer

Thanks to sci-fi, we associate artificial intelligence (AI) with robot uprisings that enslave humanity. But – for now – AI is more likely to help than harm us. Researchers have developed a new AI algorithm to help diagnose colorectal cancer, the third most common cancer worldwide. The examination of colorectal biopsies by expert pathologists is the best way to diagnose cancer, but these specialists are overworked, leading to fatigue and risking mistakes. An algorithm, however, is only limited by its hardware. The AI, trained on clinical biopsies annotated by experts, first identifies healthy and cancerous areas and then classifies each as ‘high risk’ (red in this biopsy section) or ‘low risk’ (green). The AI classified nearly 4 in 5 biopsies correctly, and most of the mistakes were false positives, meaning that it's unlikely to miss cancer. After further improvements, the aim is to introduce the AI as a tool to assist pathologists soon.

Written by Henry Stennett

Henry Stennett

Writer

Henry is a science communicator at Cancer Research UK and a freelance writer. He used to research microbiology but now he’s got the writing bug.

• Image from work by Cowan Ho, Zitong Zhao and Xiu Fen Chen, and colleagues
• Department of Anatomical Pathology, Singapore General Hospital, Singapore
• Image originally published with a Creative Commons Attribution 4.0 International (CC BY 4.0)
• Published in Scientific Reports, February 2022