Tech & Innovation in Healthcare

Question: Our practice sees several patients a year for breast ultrasounds to examine for signs of breast cancer. Some of our providers are worried about missing abnormalities with the number of patients coming in each year.

Is there a technology that could be a useful tool for examining ultrasound images?

New Hampshire Subscriber

Answer: Yes, there is a technology that could help healthcare providers when evaluating ultrasound images. Here’s one way that artificial intelligence (AI) could be the tool your radiologists are seeking.

Researchers developed a deep-learning (DL) AI model that assists radiologists while evaluating ultrasound images for signs of breast cancer. Published in Radiology: Artificial Intelligence, Volume 5, Number 5, the research describes the DL AI model as a dual attention-based convolutional neural network, which was designed to distinguish between malignant and benign breast lesions.

Researchers conducted the retrospective study at four hospitals and examined data from more than 45,000 ultrasound images from B-mode and color doppler ultrasound images. For comparison, the AI model went up against three novice readers who had less than five years of ultrasound imaging experience and two experienced readers with eight and 18 years of experience.

The DL AI model achieved an expert-level diagnostic performance with an area under the receiver operating characteristic curve (AUC) of 0.94 for the internal dataset. The external datasets AUCs were 0.92, 0.91, and 0.96 for three different hospitals, which indicates consistent performance in different settings.

“The DL model may help radiologists, especially novice readers, improve accuracy and interobserver agreement of breast tumor diagnosis using [ultrasound],” wrote the researchers.