|A new AI system can predict cancer recurrence with high accuracy using images / Photo Credit: Ravil Sayfullin via 123RF|
Artificial intelligence in healthcare plays a huge role in introducing new tools and applications to better predict and prevent diseases. These help patients address their health concerns immediately as well as understand them. A research group led by Yoichiro Yamamoto and Go Kimura from the RIKEN Center for Advanced Intelligence Project (AIP) in Japan introduced a technology that can identify features relevant to cancer prognosis that was not previously noted by pathologists.
According to Medica, the world's largest medical trade fair for medical technology, electromedical equipment, laboratory equipment, diagnostics, and pharmaceuticals, the researchers used an approach called "unsupervised learning.” AI learned medical knowledge instead of being “taught.” They used unsupervised deep neural networks, known as autoencoders so AI can learn without being given any medical knowledge. The study published in Nature Communications showed that the technology successfully found features in pathology images from human cancer patients.
This can lead to higher accuracy of predicting prostate cancer recurrence compared to pathologist-based diagnosis. "This technology can contribute to personalized medicine by making highly accurate prediction of cancer recurrence possible by acquiring new knowledge from images. It can also contribute to understanding how AI can be used safely in medicine by helping to resolve the issue of AI being seen as a 'black box’,” Yamamoto said.
The researchers acquired 13,188 whole-mount pathology slide images of the prostate from Nippon Medical School Hospital (NMSH). They discovered that the features discovered by the AI were more accurate compared to the predictions made based on the human-established cancer criteria developed by pathologists. Yamamoto stated that this new technology can help patients by allowing highly accurate predictions of cancer recurrence.
"I was very happy to discover that the AI was able to identify cancer on its own from unannotated pathology images. I was extremely surprised to see that AI found features that can be used to predict recurrence that pathologists had not identified,” Yamamoto added.