AI Shows Promise in Detecting Overlooked Breast Cancers Between Screenings, UCLA Study Reveals

Source Credit: freepik

What You Should Know: 

– Artificial intelligence (AI) could play a crucial role in identifying breast cancers that develop between routine screening mammograms, potentially detecting them before they become more advanced and challenging to treat, according to a new study led by investigators at the UCLA Health Jonsson Comprehensive Cancer Center. 

– The research suggests that integrating AI into breast cancer screening practices could lead to earlier treatment and improved patient outcomes.

The Challenge of Interval Breast Cancers

Interval breast cancers are a significant concern because they are often diagnosed at a later stage and can be more aggressive than those caught during routine mammograms. These can include tumors that were technically visible on previous mammograms but were not detected by radiologists, or those that presented with very subtle signs that were easy to miss or below the typical threshold for human detection.

The study, published in the Journal of the National Cancer Institute, found that AI was capable of identifying mammographically visible types of “interval cancers” – those diagnosed between scheduled screenings – by flagging them at the time of the initial, seemingly normal screening. Researchers estimate that this technology could help reduce the incidence of these types of breast cancers by as much as 30%.

“This finding is important because these interval cancer types could be caught earlier when the cancer is easier to treat,” said Dr. Tiffany Yu, assistant professor of radiology at the David Geffen School of Medicine at UCLA and first author of the study. “For patients, catching cancer early can make all the difference. It can lead to less aggressive treatment and improve the chances of a better outcome.”

AI’s Performance as a “Second Set of Eyes”

This retrospective study, among the first of its kind in the United States to explore AI for detecting interval cancers, analyzed nearly 185,000 past mammograms conducted between 2010 and 2019. The dataset included both digital mammography (2D) and digital breast tomosynthesis (DBT or 3D mammography), reflecting common U.S. screening practices which differ from European standards (typically 2D mammography every two to three years versus annual screening in the U.S., often with 3D).

From this data, the team focused on 148 cases where women were diagnosed with interval breast cancer. A commercially available AI software, Transpara, was then applied to the initial screening mammograms performed before the cancer diagnosis. The AI tool scored each mammogram on a cancer risk scale of 1 to 10, with a score of 8 or higher considered a flag for potential concern.

Key findings from the AI analysis included:

The AI flagged 76% of the mammograms that had originally been interpreted as normal by radiologists but were later linked to an interval breast cancer.It identified 90% of cases where the cancer had been visible on the mammogram but was initially missed or misinterpreted (missed reading errors).The AI caught approximately 89% of “minimal-signs-actionable” cancers, where very subtle signs could have reasonably prompted action.It also flagged 72% of “minimal-signs-non-actionable” cancers, those likely too subtle for a radiologist to act upon initially.For cancers that were “occult” or completely invisible on the mammogram to the human eye, the AI still flagged 69% of these cases based on the initial screening.The AI was somewhat less effective at identifying “true interval cancers” – those not present at the time of the initial screening but developed later – flagging about 50% of these.

Nuances in AI Accuracy and the Path Forward

While the results are promising, the researchers also highlighted important limitations and areas requiring further investigation. “While we had some exciting results, we also uncovered a lot of AI inaccuracy and issues that need to be further explored in real-world settings,” said Dr. Hannah Milch, assistant professor of radiology at the David Geffen School of Medicine and senior author of the study. “For example, despite being invisible on mammography, the AI tool still flagged 69% of the screening mammograms that had occult cancers. However, when we looked at the specific areas on the images that the AI marked as suspicious, the AI did not do as good of a job and only marked the actual cancer 22% of the time.”