put that theory to the test and found very promising results. Participants had their mammograms screened by either AI or two human radiologists — and AI was just as efficient at identifying signs of breast cancer.The study used over 80,000 women around the age of 54. A little over half were assigned to an AI-assisted screening, which involved at least one radiologist interpreting the results with the help of AI.
The other half were assigned to a standard mammogram screening, with one to two radiologists reading the scans themselves.The AI-assisted doctors accurately detected breast cancer in 28% of patients, while the doctors who did not use AI only spotted cancer in 25%. This meant AI assistance led to 41 more cancer diagnoses overall, including invasive and early-stage breast cancers.
Additionally, AI had about the same error rate as the radiologists: The false-positive rate was 1.5% in both groups.With the help of AI, the study concluded that fewer screen readings were needed and estimated a 44% reduction in workload.