Source: Brigham and Women’s Hospital
Summary: Investigators have developed a novel technique by taking the full advantage of artificial intelligence to detect ovarian cancer early and accurately.
Generally, ovarian cancer is diagnosed when the disease is at an advanced stage and at that point only a quarter of patients will survive for about 5 years. But for women whose cancer is fortunately picked at an early stage will have much higher survival rates. There are no FDA-approved screening techniques for ovarian cancer currently available. But the early detection tests for ovarian cancer such as protein CA125 detection, ultrasound have a high false positive rate. These test even do not have meaningful impact on patient’s survival rates. Investigators from Brigham and Women’s Hospital and Dana-Farber Cancer Institute have developed a novel technique by taking the full advantage of artificial intelligence to detect ovarian cancer early and accurately. The study was published in the journal eLife.
Principal investigators Kevin Elias, Dipanjan Chowdhury and their colleagues determined that cells of ovarian cancer and normal cells have different microRNA profiles. Normally microRNAs circulate in the blood and it is very much possible to measure their levels in the serum sample. The microRNAs from blood samples of 135 women (prior to surgery or chemotherapy) were sequenced to create a “training set” with which they trained a computer program to look out for microRNA differences between the cases of ovarian cancer, benign tumor cases, non-invasive tumors and healthy tissues. With the help of machine-learning approach, they leveraged large amounts of microRNA data which help them in developing different predictive models. The model which distinguished ovarian cancer from benign tissue with high accuracy is the neural network model, reflecting complex interactions among microRNAs.
Lead author, Kevin Elias said, “microRNAs are the copywrite editors of the genome: Before a gene gets transcribed into a protein, they modify the message, adding proofreading notes to the genome”, and further added “When we train a computer to find the best microRNA model, it’s a bit like identifying constellations in the night sky. At first, there are just lots of bright dots, but once you find a pattern, wherever you are in the world, you can pick it out.”
Senior author, Dipanjan Chowdhury said, “The key is that this test is very unlikely to misdiagnose ovarian cancer and give a positive signal when there is no malignant tumor. This is the hallmark of an effective diagnostic test.”
More Information: Kevin M Elias et al, “Diagnostic potential for a serum miRNA neural network for detection of ovarian cancer”, eLife (2017). DOI: 10.7554/eLife.28932