Deep learning neural networks now offer the best hope for detecting breast cancer. The latest CDC study reports about 247,200 new diagnoses of breast cancer in women and men and nearly 41,500 fatalities every year. Previously, we wrote about work by a research team from New York’s Rory Myers College of Nursing that showed the relative accuracy of five machine learning algorithms in detecting lymphedema in breast cancer surgery survivors. We also reported the results of a study by a research group from MIT, Harvard Medical School, and Massachusetts General Hospital (MGH) on a deep-learning model to assess dense breast tissue in mammograms.
Researchers from the Karolinska Institute in Stockholm, Sweden recently published a study in Radiology that compared the breast cancer prediction accuracy of deep learning risk scores with standard mammographic density scores. The researchers evaluated 2,283 women, 278 of whom were diagnosed with breast cancer. Using the earliest available digital mammograms for each woman with logical regression models, the deep neural network was superior to the best mammographic density model in predicting breast cancer risk. The neural network also reported fewer false negatives than the traditional model.
According to the lead author of the study, Karin Dembrower, M.D., breast radiologist and Karolinska Institute Ph.D. candidate, “The deep neural network overall was better than density-based models. And it did not have the same bias as the density-based model. Its predictive accuracy was not negatively affected by more aggressive cancer subtypes.”
The Karolinska study is another in a growing number that supports using artificial intelligence, not only in breast cancer risk assessment but in a wide range of diagnostic roles.
