AI system outperforms expert radiologists in detecting breast cancer


An artificial intelligence (AI) programme has been developed that is better at spotting breast cancer in mammograms than expert radiologists, an article published in Nature reports.

“Screening mammography aims to identify breast cancer at earlier stages of the disease, when treatment can be more successful,” lead author Scott Mayer McKinney (Google Health, Palo Alto, USA) et al write. “Despite the existence of screening programmes worldwide, the interpretation of mammograms is affected by high rates of false positives and false negatives.”

To assess the performance of their AI system in the clinical setting, the investigators curated a large, representative dataset of mammograms for 25,856 patients in the UK, and a large, enriched dataset of 3,097 patients from the USA. These datasets were used to train the AI system, before it was used to identify the presence of breast cancer in mammograms of women who were known to have had either biopsy-proven breast cancer or normal follow-up imaging results at least one year later.

In an independent study of six radiologists, the AI system outperformed all of the human readers: the area under the receiver operating characteristic curve (AUC-ROC) for the AI system was greater than that AUC-ROC for the average radiologist by an absolute margin of 11.5%. Mayer McKinney and colleagues further demonstrate an absolute reduction in false positives of 5.7% in the USA and 1.2% in the UK, and an absolute reduction in false negatives of 9.4% in the USA and 2.7% in the UK.

“This robust assessment of the AI system paves the way for clinical trials to improve the accuracy and efficiency of breast cancer screening,” Mayer McKinney et al conclude.

If the programme proves its worth in clinical trials, the software, developed by Google Health, could make breast screening more effective and ease the burden on health services such as the UK’s National Health Service (NHS), where radiologists are in short supply.

Writing in an accompanying editorial in Nature, Etta Pisano (American College of Radiology, Philadelphia, USA, and Beth Israel Lahey Medical Center, Harvard Medical School, Boston, USA) says “Screening is used to detect breast cancer early in women who have no obvious signs of the disease. This image-analysis task is challenging because cancer is often hidden or masked in mammograms by overlapping ‘dense’ breast tissue. The problem has stimulated efforts to develop computer-based AI systems to improve diagnostic performance. […] Despite some limitations, McKinney and colleagues’ study is impressive. Its strengths include the large scale of the data sets used for training and subsequently validating the AI algorithm.

“McKinney and colleagues’ results suggest that AI might someday have a role in aiding the early detection of breast cancer, but the authors rightly note that clinical trials will be needed to further assess the utility of this tool in medical practice.”


Please enter your comment!
Please enter your name here