Early and accurate diagnosis of cancer gives physicians better odds at successfully treating patients.  A new blood test uses machine learning to detect types and sites of cancer in the human body.

Recent research published in the Annals of Oncology discusses how analysis of free-floating DNA in blood samples was used to detect over 50 types of cancer of differing stages in a study of 6,689 patients.  Of these, 2,482 were known to have cancer, and 4,207 were known to be cancer free.

As noted by the study authors, cancer detection usually depends on screening for a small number of cancers, like prostate, breast, and other forms of cancers.  Oftentimes a diagnosis of less common cancers depends on presentation of symptoms or more exceptional screening efforts. Because of this, cancers may not be diagnosed until later stages when the disease is harder to treat or has spread within the body.

Patients are sometimes unable to obtain an accurate diagnosis, or the cancer is altogether missed by their treating physician—losing precious treatment time in the process.

How does the new blood test work?

Because cancer tumors shed DNA into the bloodstream, researchers believed they might be able to develop an algorithm to develop artificial intelligence (AI) that could not only detect the anomalous DNA, but to determine the original site of the tissue DNA.  Cancerous changes in the expression of DNA are called “methylation.”

Study author, Dr Geoffrey Oxnard of the Dana-Farber Cancer Institute, which is part of Harvard Medical School, said the team borrowed an approach used for pre-natal screening.  Since fetal abnormalities could be identified within a blood sample, the team felt AI could be trained to analyze shed DNA and its original location in the body.

Initial results of the trial were promising, including:

  • Approximately .07 percent of the research results were false positives, meaning that a patient without cancer could be led to believe they have cancer. Currently, standard breast cancer screening yields a false positive rate of about 10 percent.
  • The screening tool predicted the tissue within which the cancer would be found 96 percent of the time. Of those, 93 percent of the predictions were accurate.
  • The ability of the tool to predict and locate cancers increased with the severity of the disease of the patient (e.g., the higher the stage, the more likely it would be accurately detected).

Still in the development stage, researchers plan to test the tool in clinical trials for early cancer detection.  Eventually, the test could provide better detection at earlier stages with less likelihood of medical misdiagnosis.

Providing unsurpassed legal service to Maryland patients injured by medical malpractice

Schochor, Staton, Goldberg, and Cardea, P.A. has more than 35 years of successful experience representing clients and their families injured through medical negligence.  When you suffer serious injury due to a medical mistake in Baltimore, Washington, DC, or elsewhere in the US, we can help. Contact us or call 410-234-1000 to schedule a free consultation today.