“I think AI is still in its relatively early adoption phase,” Lehman said. “We have centers that are not academic centers that are very forward-thinking and really want to integrate AI into their practices. However, we are also seeing a very familiar story when we introduce computer-aided detection (CAD ) in academic and community centers.The technology is integrated into clinical care, but we are still investigating what the actual outcomes are on patients who are screened with mammography where AI tools are or are not being used.
This includes AI for automated detection of breast cancer lesions and flagging them to show areas of interest on mammography images, or to flag studies that need more attention. AI can also take a first look at mammograms to determine if they appear normal, so radiologists can prioritize which exams should be read first and which may be more complex.
This technology will likely become more important as the number of breast imaging exams shifts from traditional four-image mammograms to much larger 3D digital breast tomosynthesis exams of 50 or more images that take longer to read. AI is already being used to flag images that deserve closer examination in these datasets.
AI is also finding use as an automated way to assess breast density to help eliminate variation in the assessment of the same patient by human readers.
However, the most exciting area of AI for breast imaging is the potential for radiomics, where AI will view medical imaging in ways that human readers cannot identify very complex patterns and small ones that will help better assess patient risk scores, or what the best outcomes will be based on various cancer treatments.
“What really excites me is the area where investigators consider the power of artificial intelligence to do things that humans can’t or aren’t very good at, and then allow humans to really focus on the tasks that humans excel at today, these AI tools haven’t even really scratched the surface,” Lehman explained.
She said this area of research using radiomics goes beyond training AI to look at images like a human radiologist and instead extracting details that are typically hidden from the human eye. This includes rapid computational segmentation and analysis of disease morphology or tissue patterns seen in images, looking for minute regional structures that can be detected by AI.
“It’s not about training the AI to look at mammograms like I do, it’s about training the AI to look for patterns and signals that my human eyes and my human brain can’t detect or process. “said Lehman.
She said today that we are only scratching the surface of the data potential of AI analysis of cancers in imaging. Deep-seated patterns in cancers on imaging can tell us a lot about which concerns will or will not respond to different drugs or therapies. The AI may be able to tell us from a much deeper analysis of the imaging, including the subtypes of that particular cancer. This would allow much better adapted and personalized medicine and treatments for each patient.