How AI may help detect cancer much before traditional diagnosis methods
Even as the world tries to figure out the potential risks and threats posed by artificial intelligence (AI), a new study has shown that AI can successfully identify people at the highest risk of pancreatic cancer, using only their medical records, up to three years before actual diagnosis.
The research was done by scientists at Harvard Medical School and the University of Copenhagen, in collaboration with VA Boston Healthcare System, Dana-Farber Cancer Institute, and the Harvard TH Chan School of Public Health. The findings were published in Nature Medicine on May 8.
Pancreatic cancer is one of the deadliest cancers in the world, and its toll is projected to increase. For one thing, it is hard to detect, and two, it is expensive to screen and test for. Very often, it is detected in the advanced stages, when treatment is less effective and outcomes are dismal, the researchers said. The research findings suggest that AI-based population screening could be valuable in finding those at an elevated risk of the disease and expedite its diagnosis.
Also read: Globalisation effect: India to face a deluge of chronic diseases like cancer
Why it is significant
Currently, there are no population-based tools to screen broadly for pancreatic cancer. Those with a family history and certain genetic mutations that put them at a greater risk of pancreatic cancer are screened in a targeted fashion. But such an approach can miss cases that fall outside of those categories, the researchers said, said a report on the Harvard Medical School website.
“An AI tool that can zero in on those at highest risk for pancreatic cancer who stand to benefit most from further tests could go a long way toward improving clinical decision-making,” study co-senior investigator Chris Sander, faculty member in the Department of Systems Biology in the Blavatnik Institute at HMS, was quoted as saying.
Such an approach could speed up the detection of pancreatic cancer, lead to earlier treatment, and improve outcomes and prolong patients’ life spans, he added.
Also read: IIT-Madras researchers develop AI tool to predict cancer-causing genes
How the test was done
The researchers trained the AI algorithm on two separate data sets totalling nine million patient records from Denmark and the United States. The AI model had the task of looking for tell-tale signs of cancer risk based on the data in the records.
Based on combinations of disease codes and the timing of their occurrence, the model predicted which patients were likely to develop pancreatic cancer in the future. Notably, many of the symptoms and disease codes were not directly related to the pancreas or did not stem from it.
The researchers tested the AI model with different time scales — six months, one year, two years, and three years. They believe the model is at least as accurate in predicting disease occurrence as are current genetic sequencing tests that are usually available only for a small subset of patients.
Most importantly, the AI tool can be used on any patient whose health records and medical history are available. This is especially important, the researchers said, because many patients at high risk may not even be aware of their genetic predisposition or family history.
Also read: ICMR frames first guidelines for AI application in healthcare, research
Other similar AI tools
About 44% of people diagnosed in the early stages of pancreatic cancer survive five years after diagnosis, but only 12% of cases are diagnosed that early. Overall, cancer is a leading cause of deaths globally. Around 10 million people die of it annually, accounting for nearly one in six deaths, according to the World Health Organization. But AI may be able to reduce those figures.
Earlier this month, Fox News reported that a US hospital had seen a 23% increase in breast cancer diagnoses using AI. ProFound AI, designed to flag problem areas on mammograms, was the first technology of its kind to be cleared by the US Food and Drug Administration (FDA) in December 2018. In some cases, the AI tool has reportedly detected early signs of the disease years before a tumour would be visible on a traditional scan.
Late last month, The Guardian reported how an AI model could accurately identify cancer, which may potentially help speed up diagnosis and fast-track treatment. The AI tool designed by scientists at the Royal Marsden NHS Foundation Trust, the Institute of Cancer Research, London, and Imperial College London can accurately identify whether large lung nodules detected in CT scans are cancerous. The findings were published in the Lancet’s eBioMedicine journal.
In a similar development, scientists at the Mass General Cancer Center and the Massachusetts Institute of Technology in Cambridge have designed a new AI tool, called Sybil, to help with lung cancer screening. In has reportedly shown immense promise though the FDA is yet to clear it.
(With agency inputs)
