Home News Doctors using AI catch breast cancer more often than either does alone

Doctors using AI catch breast cancer more often than either does alone


The software being tested comes from German startup Vara, which also led the research. The company’s AI is already used in more than a quarter of breast cancer screening centers in Germany, and was introduced to one hospital in Mexico and another in Greece earlier this year.

With help from radiologists at University Hospital Essen in Germany and Memorial Sloan Kettering Cancer Center in New York, Vara’s team tested two approaches. First, the AI ​​works alone to analyze the mammogram. The AI, on the other hand, automatically distinguishes between scans that it thinks look normal and those that cause concern. It recommends the latter to radiologists, who review the AI’s assessment before seeing it. Then, if the doctor doesn’t detect cancer, the AI ​​will issue a warning.

“In the proposed AI-driven process, nearly three-quarters of screening studies do not require radiologist review, while improving overall accuracy.”

Charles Langlotz

To train the neural network, Vara provided AI data from more than 367,000 mammograms—including radiologists’ notes, raw assessments, and information about whether a patient ended up with cancer—to learn how to put those scans into One of three buckets: “Confident Normal,” “Unconfident” (no predictions given), and “Confident Cancer.” The conclusions of the two methods were then compared to the decisions made by real radiologists initially on 82,851 mammograms from screening centers that did not provide scans for training the AI.

The second method — doctors and AI working together — was 3.6 percent better at detecting breast cancer than doctors working alone, and produced fewer false positives. It did this while automatically shelving scans it was classified as confidently normal, accounting for 63% of all mammograms. This intense streamlining can reduce the workload for radiologists.

After breast cancer screening, patients with normal scans are sent on the road, while patients with abnormal or unclear scans trigger follow-up testing. But radiologists who review mammograms miss one in eight cancers. Fatigue, overwork, and even the time of day can affect a radiologist’s ability to identify tumors as they review thousands of scans. Visually subtle signs are also generally less likely to set off alarms, and dense breast tissue, found mostly in younger patients, makes cancer signs harder to spot.

German law requires radiologists who use AI in the real world to view every mammogram, at least those that the AI ​​calls normal. While radiologists can always turn down calls from AI, AI can still help them by pre-filling scan reports marked as normal.

Thilo Töllner, a radiologist and head of the German Breast Cancer Screening Center, has been using the procedure for two years. He sometimes disagrees when the AI ​​categorizes scans as confidently normal and fills out reports manually to reflect different conclusions, but says “normal is almost always normal”. Most of the time, “you just need to hit enter.”

Mammograms flagged by the AI ​​as equivocal or “confirmed cancer” are referred to radiologists — but only after the doctor provides an initial independent assessment.

Source link

Previous articleCelebrity Cruises Quietly Makes Changes To A Few Captains Club Loyalty Program Features
Next articleShe’s too eager to meet my celebrity friends


Please enter your comment!
Please enter your name here