Artificial Intelligence Offers New Way Of Detecting Breast Cancer
Researchers believe that artificial intelligence offers a new way to detect benign and high-risk breast lesions to eliminate unnecessary surgery. A study published online in the journal Radiology indicates the potential use of AI technology in the future to improve the treatment of breast cancer patients.
Breast Cancer Risks To Women
About 40,000 women die in the United States each year because of breast cancer. In some cases, the disease can be cured when detected earlier. The mammogram is one of the tests available to recognize breast cancer, but it often leaves a false positive result leading to biopsies and surgeries.
The most common cause of a false positive result is high-risk lesions. The lesions may seem to be suspicious in the mammogram with abnormal cells on the needle biopsy test. The patient would then undergo a surgery to eliminate the lesion, but some of the high-risk lesions appear to be 90 percent benign. This procedure puts women at risk every year as they need to go through unnecessary and expensive surgeries.
New Way To Detect Breast Cancer
MIT's Computer Science and Artificial Intelligence Laboratory, Massachusetts General Hospital, and Harvard Medical School researchers believe that artificial intelligence is the answer to diminish unnecessary surgery while maintaining the efficiency of mammogram in detecting breast cancer.
The teams worked together to improve detection and diagnosis by developing an AI system that utilizes machine learning. The model has approximately 600 existing high-risk lesions and checks for patterns of different data elements to analyze the traditional risk factors including family history, pathology reports, past biopsies, and demographics.
Researchers tested the machine learning model on 335 lesions, and it correctly predicted 97 percent of breast cancer diagnosis as malignant and decreased almost 30 percent of benign surgeries.
"To our knowledge, this is the first study to apply machine learning to the task of distinguishing high-risk lesions that need surgery from those that don't," says Professor Constance Lehman, Harvard Medical School, who is also chief of the Breast Imaging Division at MGH's Department of Radiology.
The Use of Machine Learning Model In The Future
"In future work we hope to incorporate the actual images from the mammograms and images of the pathology slides, as well as more extensive patient information from medical records," says Manisha Bahl, Massachusetts General Hospital Breast Imaging Fellowship Program director.
This is the first step for the medical community to embrace machine learning in recognizing patterns and trends that are invisible to humans. MGH radiologists will incorporate the machine learning model into clinical practice over the next year as they are still working to further hone the model.