Mammograms have been a crucial tool in preventive screening for breast cancer. Still, data suggests about 1 in 8 breast cancers are missed through false negative screening.
Dr. Christoph Lee is co-leading a clinical trial at University of Wisconsin and six other academic research centers in the U.S. to validate the use of an artificial intelligence platform as a new tool to help radiologists interpret mammograms. Their goal is to improve accuracy in mammogram screening across all health care settings and help diagnose breast cancer in its earliest, most treatable stages.
“The whole goal of routine screening for breast cancer is to detect them at the earliest stages, when they can be cured, so if we can detect cancers earlier with AI, then we can start treatment sooner,” said Lee, a UW Health | Carbone Cancer Center researcher and professor of radiology at the University of Wisconsin School of Medicine and Public Health.
People are also reading…
As part of this, Lee is leveraging the statewide connections built through Carbone Cancer Center’s Wisconsin Oncology Network of Imaging Excellence (WONIX). This cooperative network offers participating Wisconsin medical clinics access to Carbone resources, such as cutting-edge imaging technology and clinical trial opportunities. The network also establishes valuable information sharing Carbone can use to advance new research.
The clinical trial Lee co-leads involves observing whether one of the leading FDA-approved AI tools for 3D mammography interpretation can help radiologists detect more cancers when used in the clinic.
All patients will receive the same mammography exam, and the results will be interpreted by a radiologist. A randomized selection of exams will also include use of the AI tool to provide an extra review that could flag potential suspicious lesions that a radiologist may miss when looking with just their eyes alone.
“We will observe through usual practice how patients do when their radiologists interpret screening mammograms with or without the use of AI assistance,” Lee explained. “At the end of the day, the radiologists will always make the final call (on the results and next steps). This isn’t an AI versus radiologist’s situation. It’s a question of, ‘Can this AI tool help a radiologist interpret mammograms better, or not?’”
The AI review is being applied randomly among patients enrolled in the study to ensure a true measure of whether it is enhancing accuracy and improving diagnoses.
To expand trial access beyond academic research centers, Lee said the Wisconsin Oncology Network of Imaging Excellence has been a key resource. They want to explore AI’s potential as a tool for smaller health care centers without sub-specialized radiologists who serve a more rural population.
“If we can show these AI tools improve care and are relatively easily implemented, it’s an advanced technology that can start being offered to benefit all women, including those in rural care settings,” he said.
