A remarkable study in the journal Nature was published this week describing how artificial intelligence can assist radiologists in comparing consecutive past mammograms from 1 to 10 years ago and predicting the development of breast cancer. The researchers used an artificial intelligence technique called “deep learning” in assessing 171,000 mammograms from 9,000 women. Their findings were remarkable.

By comparing year-to-year mammograms, they found that:

  1. Subtle changes in breast density through the years could predict breast cancer (dense breasts are the radiologist’s nightmare, but 34% of women have dense breasts, making it extremely common)
  2. Using this technique, the results were divided into percentiles. A deep learning result above the 90th percentile showed breast cancer developing in 38%. Conversely, those in the lower 10th percentile had only a 0.47% incidence of breast cancer developing.
  3. In women with a history of breast cancer, it was able to predict a recurrence of their cancer not only in the same breast, but the opposite breast as well.
  4. Early breast cancer, which is usually not identified on a single mammogram, can now be identified, resulting in an earlier biopsy, hence an earlier diagnosis with a higher cure rate.

Why this matters?

  1. Patients with a mammogram resulting in a suspicious area, but not quite worrisome enough to warrant a biopsy, could be evaluated by more frequent mammograms, ultrasound and/or MRI.
  2. Mammograms showing stability through the years can be used as a guide to less frequent mammograms, lowering radiation exposure, and lowering the frequency of biopsies.
  3. To emphasize what was stated above, this technique can diagnose breast cancer earlier, which leads to higher cure rates, more conservative surgery, and less toxic chemotherapy.

Two things to keep in mind:

1. This technique is not yet commercially available (the study was done in the Netherlands)

2. Although the study was conducted in 9,000 women with 171,000 mammograms, this sample size is relatively small. Its true utility will only be proven in a larger study involving hundreds of thousands of mammograms in multiple diverse populations.

Dr. Charlie Barnett is a contributor to KnoxTNToday, where he writes a weekly column, DocTalk, sharing his expertise on health and wellness management.

Follow KnoxTNToday on Facebook and Instagram.   Get all KnoxTNToday articles in one place with our Free Newsletter.