AI Assists Doctors in Identifying More Cases of Breast Cancer in the Largest Real-World Study

AI Assists Doctors in Identifying More Cases of Breast Cancer in the Largest Real-World Study

Researchers reported on January 7 in Nature Medicine that AI-assisted mammogram analysis helped doctors detect one additional cancer case for every 1,000 women screened compared to when the technology wasn’t used. This finding comes from the largest real-world study on AI’s role in breast cancer screening, which involved nearly 500,000 women in Germany. The results suggest that AI could enhance the screening process without increasing the number of false positives.
AI rivals doctors’ ability to interpret mammograms, a real-world study with nearly 500,000 participants in Germany suggests. Credit: Depositphotos

Researchers reported on January 7 in Nature Medicine that AI-assisted mammogram analysis helped doctors detect one additional cancer case for every 1,000 women screened compared to when the technology wasn’t used. This finding comes from the largest real-world study on AI’s role in breast cancer screening, which involved nearly 500,000 women in Germany. The results suggest that AI could enhance the screening process without increasing the number of false positives.

AI in mammography screening is at least as effective as a human reader, and our study shows it’s even better,” said cancer epidemiologist Alexander Katalinic from the University of Lübeck in Germany.

Germany’s Dual Radiologist Mammogram Review Process

Germany’s breast cancer screening program mandates that two radiologists independently review each patient’s mammograms to identify spots, abnormal masses, and other irregularities. (In the U.S., most clinics typically use just one physician.) If at least one radiologist suspects cancer based on the four X-ray images, which are compared to the patient’s previous screenings, a third radiologist is brought in to determine if further tests are needed.

We have 3 million women participating in this program each year, and 24 million images need to be reviewed annually,” says Katalinic. “That’s a substantial workload for the radiologists.”

In the new software, an AI-supported image viewer triggers a “safety net” alert if a radiologist deems a patient’s mammograms as cancer-free but the AI software suspects otherwise. The white square shows the suspicious region.

AI-Assisted Mammogram Review

To determine if AI could ease the workload, decision referral software was introduced at 12 screening sites nationwide. Over 460,000 women aged 50 to 69 participated in the study from July 2021 to February 2023. The AI software categorized the mammograms as normal, suspicious, or unclassified. The 119 participating radiologists opted to use an AI-assisted image viewer, which displayed the software’s assessments, for about half of the screenings.

Without AI, clinicians identified roughly six confirmed breast cancer cases per 1,000 patients during screening. With AI’s help, they detected around seven cases, resulting in a 17.6 percent higher detection rate. The group screened with AI had slightly fewer false positives compared to those receiving traditional screenings, where cancer is suspected but ruled out with further tests.

Although the ideal way to integrate AI into radiologists’ workflows is still unclear, Stefan Bunk, CTO and cofounder of Vara, the Berlin-based healthcare tech company behind the AI, suggests that it could replace one of the initial readers. “This is a conversation that should now begin,” he says.


Read the original article on: Science News

Read more: Discovery of Cancer Protein Uncovers New Treatment Target

Share this post

Leave a Reply