A recent article from The Guardian discussed a new blood test that screens for hard-to-detect cancers. The system, developed at Harvard Medical School’s Dana-Farber Cancer Institute, not only detects cancer, but also uses machine learning to indicate what type of cancer the patient has. The team behind the concept gave the system data related to methylation patterns in DNA from blood samples of over 2,800 patients. Roughly half of the sample group had cancer, while the other half did not.
Results showed that less than 1% of the patients without cancer were wrongly identified by the system. The test was able to offer a prediction for where the cancer originated with 93% accuracy. The system is now being explored in clinical trials.