Dimitrios Mathios, MD, assistant professor of neurosurgery at Washu Medicine, has devised a method of using machine learning to detect brain tumors by analyzing cancer-related DNA patterns in the blood. In a recent study, Mathios found that the machine learning technology detected cancer in 73% of cases from a cohort drawn from patients in the US and South Korea. Results were validated in a second cohort of samples from brain cancer patients in Poland. In contrast, previous studies found that conventional fluid biopsies testing blood with standard cancer biomarkers were less than 10% of brain tumors.
Mathios' research was published in Cancer Discovery on April 29th.
Early detection of brain tumors before they grow and spread can improve treatment outcomes. However, early detection of brain cancer has proven challenging to prevent blood-brain barriers (prevent harmful substances in the bloodstream from entering the brain) as they prevent specific biomarkers that reveal the presence of brain tumors invade the circulation system.
The presence of brain tumors is difficult to detect without expensive scans. This may not be ordered before the cancer progresses to the point where it is already causing detectable symptoms. However, even early in tumor growth, disease often causes altered body immune responses. Tumor cells can also throw DNA fragments into the bloodstream, which have properties that differ from those released from normal tissue. By searching for patterns consistent with these effects in blood samples, the machine learning tool developed by Mathios and his collaborators was able to reliably detect brain cancer without invasive or expensive procedures.
Researchers estimated that the widespread adoption of the tool for screening individuals with symptoms such as headaches could potentially detect as early as 1,700 additional brain tumor cases in the United States per year.
