AI Tech Sees Prostate Cancer Humans Cannot
A Second Look with Artificial Intelligence
Even when a pathologist gives a clean bill of health, an early form of prostate cancer might be hiding in plain sight. A groundbreaking study from Uppsala University has revealed that Artificial Intelligence can detect subtle tissue changes that signal cancer long before they are visible to the human eye, offering new hope for early detection.
Previous research has already established AI's potential in identifying cancer-related tissue changes. However, this latest study, published in Scientific Reports, takes it a step further by showing that AI can successfully find cancers that were initially missed by expert pathologists.
The 'Missed Study': A Groundbreaking Approach
The project, aptly nicknamed the ‘missed study,’ focused on tissue samples that pathologists had originally cleared. The results were startling.
“We have now shown that with the help of AI, it is possible to find signs of prostate cancer that were not observed by pathologists in more than 80 per cent of samples from men who later developed cancer,” explains Carolina Wählby, Professor of Quantitative Microscopy at the Department of Information Technology and SciLifeLab, who was instrumental in the AI's development.
The research, a collaboration with Umeå University, analyzed samples from 232 men, all of whom were initially deemed healthy. Within just two and a half years, half of these men developed aggressive prostate cancer. The other half remained cancer-free for at least eight years.
Training AI to See the Invisible
Since all the initial tissue samples were assessed as negative, the research team pioneered a new method for training their AI tool. The AI was trained to analyze every biopsy image piece by piece. It operated on the assumption that the biopsies from men who later got sick must contain abnormal patterns, while the others would not. The AI was then validated on a separate, independent set of images.
“When we looked at the patterns that the AI ranked as informative, we saw changes in the tissue surrounding the glands in the prostate – changes also observed in other studies,” says Wählby. “This shows that AI analysis of routine biopsies can detect subtle signs indicating clinically significant prostate cancer before it becomes obvious to a pathologist.”
The Future of Cancer Screening
The researchers propose that this type of AI analysis could become a vital tool in determining how frequently men with negative biopsy results should be monitored. By identifying those with subtle but high-risk tissue changes, clinicians can provide more personalized follow-up care. In a commitment to advancing the field, the imaging data and the methods developed by the researchers have been made openly available for further development and research.
