Medical

AI to help identify breast cancer risk and reduce unnecessary surgeries

AI to help identify breast cancer risk and reduce unnecessary surgeries
One in five breast cancers are not picked up by trained radiologists looking at mammograms, but AI can help improve this result
One in five breast cancers are not picked up by trained radiologists looking at mammograms, but AI can help improve this result
View 1 Image
One in five breast cancers are not picked up by trained radiologists looking at mammograms, but AI can help improve this result
1/1
One in five breast cancers are not picked up by trained radiologists looking at mammograms, but AI can help improve this result

Every year thousands of woman undergo painful and invasive surgeries to remove breast lesions that current diagnostic tools identify as of a high risk for cancer. The vast majority of these procedures reveal the lesions to be benign, so improving current detection and diagnosis tools is a major priority for many researchers. Now an AI system that uses machine learning has been developed to predict which high-risk lesions are most likely to become cancerous.

Mammograms are still the most important diagnostic tool for uncovering breast cancer. Suspicions lesions that are discovered through a mammogram will subsequently be tested with a needle biopsy. Generally if that biopsy is found to be abnormal a patient will undergo surgery to remove the lesion, but 90 percent of the time these lesions are found to be benign rendering the surgical procedure unnecessary.

"Because diagnostic tools are inexact, there is an understandable tendency for doctors to over-screen for breast cancer," says coauthor of the new research Regina Barzilay. "When there's this much uncertainty in data, machine learning is exactly the tool that we need to improve detection and prevent overtreatment."

The research team from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL), Massachusetts General Hospital and Harvard Medical School developed a machine learning model that was trained on 600 existing high risk lesions, incorporating variables as broad as family history, demographics and past biopsies. The model was then tested on 335 lesions, correctly predicting with an accuracy of 97 percent, those that were ultimately upgraded to cancer.

The study concluded that over 30 percent of surgeries to remove these benign lesions could have been avoided by incorporating this machine learning model into general diagnostic practices.

"In the past we might have recommended that all high-risk lesions be surgically excised," says Constance Lehman, one of the collaborators on the project. "But now, if the model determines that the lesion has a very low chance of being cancerous in a specific patient, we can have a more informed discussion with our patient about her options. It may be reasonable for some patients to have their lesions followed with imaging rather than surgically excised."

AI diagnostic tools are booming in medical research at the moment, with machines helping doctors scour through data to better diagnose a variety of conditions. From a Chinese system helping focus lung cancer diagnosis to an AI that can identify skin cancers as well as a trained dermatologist, the ability for machine models to help identify deadly disease is increasing from year to year.

The new study was published in the journal Radiology.

Source: MIT Computer Science and Artificial Intelligence Laboratory

No comments
0 comments
There are no comments. Be the first!