Though much of artificial intelligence in 2024 has been focused on improving existing systems and technologies, there are plenty of companies and researchers testing what large-language models are capable of, and how they might impact other critical industries, from transportation to finance - and, as seen in a recent study from the University of Copenhagen, the medical industry.
Using deep learning technology developed at the university, researchers analyzed mammary tissue biopsies from donors to look for signs of damaged cells - an indicator of cancer risk, particularly breast cancer (which caused 670,000 deaths worldwide in 2022 alone).
What they found was that the AI technology, trained on cells developed and intentionally damaged in cell culture, was far better at predicting risk than typical clinical benchmarks for cancer assessment. The trained AI would then search for “senescent” cells (active cells that are no longer dividing) in tissue biopsies. By using two LLMs (or one model with the “Gail score”, the industry’s current standard model), they got results far better at predicting breast cancer risk than just the Gail model alone.
Researchers hope the information will allow for improved screening and treatment protocols, identifying high-risk clients and reducing the burden of regular biopsies for lower-risk patients - though it may be several years before the technology is available to use at hospitals and clinics