A key ally in our fight against cancer is early detection. The sooner you discover a potential tumor, the sooner you can take action, and the likelier it is that a full recovery is made.

Unfortunately, cancer can be insidious. It can brew up for years without causing clear symptoms, and diagnostic tests can be painful, expensive, and intrusive. That might change soon.

In a new study, researchers from the Dana-Farber Cancer Institute and Harvard Medical School presented the results of a new blood test for cancer. The diagnosis was tested against 4,000 samples from patients, some of which had cancer, and some of which didn’t.

Remarkably, the test proved accurate at detecting over 50 types of cancer, including bladder, esophagus, lung, and breast cancer. The test had a 98.3% to 99.8% accuracy for positive results and a 0.7% false-positive rate — not perfect, but remarkably accurate. In the samples where cancer was detected, the test was also able to pinpoint its type of cancer in 93% of instances (in the others, it correctly identified the presence of cancer, but misrepresented the type).

The test looks for chemical markers in the bloodstream (methylation patterns in DNA). These bits of cell-free, free-floating DNA pieces are leaked from tumors into the bloodstream and can be a tell-tale indicator. The team used a machine learning algorithm (a type of artificial intelligence) to train the test to look for patterns that indicate the presence of cancer and classify it accordingly.

It’s still in its early days, and it’s unclear how the test will perform in a broader sample size — particularly one where no information is known about the patients. Researchers plan to further improve the test’s detection rate — particularly in the early stages, where detection rates were substantially lower. Detecting cancers at their earliest stages, when they are less aggressive and far more treatable, is a key objective for the test.

The fact that the test is painless and scalable is promising, but it remains to be seen whether its efficacy will be confirmed in larger trials. A blood-based multi-cancer detection test should demonstrate certain fundamental performance to be useful in a general screening population.

Now, researchers are already embarked on several clinical trials to test the validity of the new diagnosis. It will be a while before such a test will become readily available, but so far at least, things are looking promising.

The study has been published in the journal Annals of Oncology.