New AI detects breast cancer from ultrasounds

Artificial intelligence (AI) machine learning is rapidly transforming how physicians, clinicians, pathologists, and health care providers diagnose patient conditions. A recent NYU Langone Health study published in Nature Communications shows how AI applied to ultrasound images can identify breast cancer with radiologist-level accuracy, reduce requested biopsies by 27.8 percent, and significantly decrease false positive rates of breast cancer by 37 percent.

“In this work, we present an AI system that achieves radiologist-level accuracy in identifying breast cancer in ultrasound images,” wrote Krzysztof Geras, PhD., the study senior investigator and assistant professor at NYU Grossman School of Medicine, in collaboration with co-investigator and radiologist Linda Moy, MD. a professor at NYU Grossman School of Medicine, and their research colleagues. Both Geras and Moy are members of the Perlmutter Cancer Center.

Breast cancer is a leading cause of death among women worldwide. Globally last year there were around two million new breast cancer cases according to figures from Susan G. Komen.

According to estimates from the American Cancer Society (ACS), breast cancer is one of the most common causes of cancer in American women. There is a 1 in 8 chance that an American woman will develop breast cancer, and a 1 in 39 chance of dying from breast cancer.

Additionally, breast cancer is not just limited to women. In 2021 there will be 2,650 new cases of invasive breast cancer and 530 breast cancer deaths in American men according to the estimates from Susan G. Komen.

The mammogram is the most commonly used imaging method for breast cancer screening and detection. However, the accuracy drops significantly from 85 percent to only 48-64 percent in cases where the person has dense breast tissue. Thus, those with dense breasts have an increased risk of four times greater of developing breast cancer.

Another drawback to mammography is that it requires expensive specialty equipment, skilled operators, and radiologists—something that is out of reach for remote and rural areas where health care specialist resources and budgets are limited.

“Given the limitations of mammography, ultrasound (US) plays an important role in breast cancer diagnosis,” the researchers wrote. “It often serves as a supplementary modality to mammography in screening settings and as the primary imaging modality in many diagnostic settings, including the evaluation of palpable breast abnormalities.”

However, ultrasound is not without its challenges. “Though consistently shown to detect mammographically occult cancers, breast ultrasound has been noted to have high false-positive rates,” the NYU Langone Health researchers wrote.

In this study, the researchers filtered a dataset of over 8.4 million images from more than 212,700 unique patients, from over 425,500 U.S. breast exams, and used it to train and evaluate an artificial neural network. The training dataset use had over 3.9 million images from over 209,100 exams from more than 101,400 patients.

The AI deep learning algorithm used was a convolutional neural network (CNN). The deep neural networks were developed in PyTorch. According to the researchers, the AI scored an area under the receiver operating characteristic curve (AUROC) of 0.976.

“With the help of the AI, radiologists decrease their false positive rates by 37.3% and reduce requested biopsies by 27.8%, while maintaining the same level of sensitivity,” reported the scientists. “This highlights the potential of AI in improving the accuracy, consistency, and efficiency of breast ultrasound diagnosis.” Psychology Today