Machine Learning Model Predicts Risk of Upgrade to Breast CA

Postado Outubro 19, 2017

TUESDAY, Oct. 17, 2017 (HealthDay News) - A machine learning model can predict the risk of upgrade of high-risk breast lesions (HRLs) to cancer, according to a study published online Oct. 17 in Radiology.

"This work highlights an example of using cutting-edge machine learning technology to avoid unnecessary surgery", says Marc Kohli, director of clinical informatics in the Department of Radiology and Biomedical Imaging at the University of California at San Francisco.

Using machine learning, researchers tested 335 high-risk lesions and correctly diagnosed 97 percent of the breast cancers as malignant, reducing the number of benign surgeries by more than 30 percent. Surgical removal is typically the recommended treatment option for these lesions due to the increased risk, even though many of these lesions do not pose an immediate threat. Some do surgery in all cases, while others perform surgery only for lesions that have higher cancer rates, such as "atypical ductal hyperplasia" (ADH) or a "lobular carcinoma in situ" (LCIS). "This is the first step toward the medical community embracing machine learning as a way to identify patterns and trends that are otherwise invisible to humans".

Doctors manage high-risk lesions in different ways.

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.

A machine learning model can predict the risk of upgrade of high-risk breast lesions to cancer, according to a study published online October 17 in Radiology.

Researchers developed a machine learning model that analyzes traditional risk factors, such as patient age and lesion histology, as well as words that appear in the text of biopsy pathology reports.

Manisha Bahl, M.D., M.P.H., from Massachusetts General Hospital in Boston, and colleagues identified consecutive patients with biopsy-proven HRLs who underwent surgery or at least two years of imaging follow-up.

"In the near future, we'll be having the data run through our machine learning model so we can get the percentage of the risk that a particular lesion would or would not upgrade to cancer if we sent them to surgery", she adds.

"Use of this model could decrease unnecessary surgery by almost one-third and could help guide clinical decision making with regard to surveillance versus surgical excision of HRLs", the authors write. The machine-learning model also found that the keyword terms "severely" and "severely atypical" in pathology reports were associated with a greater risk of developing cancer.