this is cool
Scientists at the Massachusetts Institute of Technology are fighting to make breast cancer diagnosis more efficient — and they've turned to artificial intelligence to do so.
Traditionally, women undergo regular mammograms, which provide images of the breasts that doctors use to identify any lesions. But while mammograms can categorize lesions as "high risk," they cannot do so with foolproof accuracy, and a needle biopsy must be performed to determine whether the tissue is in fact cancerous. Ninety percent of these lesions are determined to be non-cancerous, MIT notes, but only after the invasive procedure has been performed.
That's where the AI comes in. Researchers at MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL), together with Massachusetts General Hospital in Boston, developed a groundbreaking new model that uses machine learning to evaluate high-risk lesions before surgery. The model, known as a "random-forest classifier," is armed with information about more than 600 existing cases, and it uses that information to identify patterns across different data points, including demographics and medical history, to more accurately predict whether lesions will become cancerous without performing the biopsy.
Additionally, some doctors perform surgery in all cases of high-risk lesions, while others look only for specific types of lesions that are known to have a higher chance of becoming cancerous before operating. The team's model yielded more accurate diagnoses despite screening for more cancers, correctly diagnosing 97 percent of cancers, MIT said, as opposed to just 79 percent via surgery on traditional high-risk lesions.
Because the traditional diagnostic tools, like mammograms, are "so inexact," doctors tend to over-screen for breast cancer, said MIT CSAIL professor Regina Barzilay, a lead author on the study and recent MacArthur "genius grant" winner. That leads to the unnecessary, expensive surgeries that find legions to be benign. "A model like this ... hopefully will enable us to start to go beyond a one-size-fits-all approach to medical diagnosis," Barzilay said.