United States: According to new reports, a new artificial intelligence foundation model could accurately detect a number of cancer types, estimate all possibilities of treatment, and provide a prognosis of expected survival, as the latest advance in medical diagnosis brought by the technology.
More about the news
“Chief” is the model created by scientists at Harvard Medical School. It is rather noteworthy due to its capability to predict the number of tumor outcomes of patients, as stated by the authors.
The Chief shows how AI has enhanced diagnosis using even pictures because it can see the importance of aspects that even the human eye may not notice.
More about the latest invention
According to Kun-Hsing Yu, assistant professor of biomedical informatics at Harvard Medical School’s Blavatnik Institute, “Our ambition was to create a nimble, versatile ChatGPT-like AI platform that can perform a broad range of cancer evaluation tasks,” as ft.com reported.
“Our model turned out to be very useful across multiple tasks related to cancer detection, prognosis, and treatment response across multiple cancers,” he noted.
Though the recent advances in AI technology have raised concerns over the misuse of advanced technology, optimists have noted that AI holds the potential to yield beneficial impacts to human society in areas such as medicine and climatology.
Chief, detailed in a paper released in Nature today, operates by analyzing digital slides of cancerous tissues. They trained it on 15mn unlabelled sections of images and then on 60,000 whole slide images or WSI of tissues containing 19 types of cancer.
According to the scientists, the plan was to ensure that Chief could link specific changes in one of the regions of the tissue to the overall picture. It was evaluated on almost twenty thousand whole-slide images from twenty-four different hospitals and patient groups all over the world.
Chief outperformed other AI diagnostic methods in cancer cell detection, the prognosis of patients’ outcomes, tumor localization, and the identification of genetic patterns for treatment response by up to 36 percent, the paper stated, as ft.com reported.
This allowed it to retain its performance despite the various methods that could be used to collect and scan the tumor cells as opposed to some other models currently, said the researchers.
Chief demonstrated a total cancer accuracy rate of about 94 percent, which rises to 96 percent for the esophagus, stomach, colon, and prostate carcinoma.
Additionally, Yu added that if Chief and the same approach are approved for future research, then they could be used to “identify early on cancer patients who may benefit from experimental treatments targeting certain molecular variations”, which would include the countries where that has not currently been done.