“I think there will be a significant evolution of the imaging platform over time to account for those differences and to make it a better presentation that allows a quick overlook of the breast for density assessment and also lesion detection, where you look at the breast globally for asymmetries and changes over time,” she said.

The advent of AI is crucial to radiologists who foresee it as a point-of-reference for confirming their interpretation of findings, and as an assistant that can read images and reports at a greater pace, enabling quicker diagnoses and the onset of earlier treatments.

Rodney Hawkins, vice president of marketing for cancer detection at iCAD, believes this potential will further evolve with the integration of deep-learning and neural networks, enabling algorithms to learn for themselves rather than being taught by radiologists.

“We can train algorithms on thousands of cancer cases or breast density cases. You can show it breast images of different densities and those images get annotated by a radiologist,” he said. “You can tell it this is category A, B, C, or D, and then that algorithm learns from the annotated images what falls into what category. It’s able to learn much faster by showing it the images.”

AI is expected to transform imaging modalities like automated ultrasound systems and breast MR. Luke Delaney, general manager of automated breast ultrasound at GE Healthcare, said that in addition to improving breast density assessment, machine learning and other advancements could enable automated devices such as GE’s ABUS to tailor care on an individual level.

“One of the things we’re working on with customers and patients is the right combinations of technologies that work together to find as many cancers as possible and keep recall rates and false positives low,” he said. “How they should be used together for each patient and her own profile is one of the big areas of research as well.”

Experts agree that in order to capitalize on these potential capabilities, greater access to data is needed. Providers can help in this endeavor by retaining and utilizing all information at their disposal, including raw data of mammographic images, which often are disposed of and not stored in PACS so hospitals and clinics won’t have to double storage capacity.

“Not having access to historical mammograms limits the capacity to compare breast density between current and prior mammograms,” said Mohamed Abdolell, CEO of Densitas Inc. “Integration directly with FFDM scanners is required when processing raw images, adding integration burden and costs, and typically requiring more effort, time and complexity of ongoing support.”

More Magazine Features

Women's Health Homepage