Dive Brief:
- Machine learning can detect liver cancer by analyzing changes in cell-free DNA fragments in the blood, according to a paper published in Cancer Discovery.
- The study applied a machine-learning model to samples from 724 people. The system performed best in people at average risk of liver cancer, achieving a sensitivity of 88% and specificity of 98%.
- Based on the results, the researchers claim the blood test can double the number of liver cancer cases detected, compared to the standard blood test, and increase early cancer detection.
Dive Insight:
Researchers at Johns Hopkins Kimmel Cancer Center in recent years have worked on non-invasive ways of assessing cell-free DNA to detect tumors, leading to a 2021 paper on the use of the technology in lung cancer. Now, the scientists have used their DELFI system, an acronym for DNA evaluation of fragments for early interception, to detect liver cancer in blood samples.
DELFI works by looking at the size and amount of cell-free DNA present in the blood from different regions across the genome. Cancer cells release DNA fragments into the blood when they die, presenting an opportunity to detect tumors early and without invasive tests.
The liver cancer study evaluated samples from 724 people from the U.S., the European Union and Hong Kong who either had liver cancer or were at average or high risk of developing the disease. DELFI achieved 88% sensitivity and 98% specificity in the average-risk group, and 85% sensitivity and 80% specificity in the high-risk cohort.
Amy Kim, M.D., assistant professor of medicine at the Johns Hopkins University School of Medicine, set out the potential implications of the work.
“Currently, less than 20% of the high-risk population get screened for liver cancer due to accessibility and suboptimal test performance, Kim, co-senior author on the study, said in a statement on John Hopkin’s website. “This new blood test can double the number of liver cancer cases detected, compared to the standard blood test available, and increase early cancer detection.”
