Artificial intelligence enables more accurate stratification of breast cancer risk
An artificial intelligence model trained solely on mammography images demonstrates greater accuracy than breast density in estimating five-year breast cancer risk, paving the way for more personalized and earlier screening.
consalud.es
Breast cancer risk prediction could undergo a significant change thanks to artificial intelligence. A new study presented at the annual meeting of the Radiological Society of North America (RSNA) shows that an AI model based solely on images achieves a more robust and accurate five-year risk stratification than traditional breast density assessment.
The research calls into question the usual methods of risk estimation, which rely on factors such as age, family history, genetics, and breast density itself. The study’s lead author, Constance D. Lehman, MD, professor of radiology at Harvard Medical School, highlights the limitations of these approaches. “More than two million women are diagnosed with breast cancer each year, and for most, it comes as a complete surprise,” she said. “Only 5% to 10% of breast cancer cases are considered hereditary, and breast density alone is a very weak indicator of risk,” she said.
The model analyzed, Clairity Breast, is the first breast cancer risk system based exclusively on imaging that has received FDA approval. Its development involved 421,499 mammograms from 27 centers in Europe, South America, and the United States, including images of women who developed cancer and those who did not in the following five years.
“The model is capable of detecting changes in breast tissue that the human eye cannot see.”
This approach enabled the AI to learn subtle patterns and differences in breast tissue associated with future risk. The model was then calibrated with an independent test set using a deep convolutional neural network capable of generating five-year risk probabilities.
“The model is capable of detecting changes in breast tissue that the human eye cannot see,” says Dr. Lehman. “This is a task that radiologists simply cannot perform. It is a task distinct from detection and diagnosis, and it will open up a whole new field of medicine, harnessing the power of AI and the untapped information in the image.”
To validate its performance, the system was applied to 236,422 bilateral two-dimensional screening mammograms from five US centers and 8,810 from a European center, obtained between 2011 and 2017. The researchers collected breast density reported by radiologists and five-year cancer outcomes from medical and tumor registries. The risks estimated by AI were classified according to the National Comprehensive Cancer Network thresholds as low, intermediate, and high risk. Using statistical models that took follow-up time into account, these categories were compared with breast density.
The results showed a clear difference. Considering breast density, women classified as high risk by AI had a cancer incidence more than four times higher than women at medium risk. In contrast, breast density alone allowed only a modest separation between groups.
“The results of this large-scale analysis demonstrate that AI risk models provide a much more robust and accurate risk stratification for five-year cancer prediction than breast density alone,” said Christiane Kuhl, MD, director of the Department of Diagnostic and Interventional Radiology at RWTH University Hospital in Aachen. “Our findings support the use of image-based AI alone as a complement to traditional markers, favoring a more personalized approach to screening.”
“Our findings support the use of image-based AI alone as a complement to traditional markers.”
Currently, the American Cancer Society recommends that women at average risk begin annual mammograms at age 40. However, breast cancer diagnoses are increasing more rapidly in women under 40, who also have more cases of advanced disease.
“An AI image-based risk score can help us identify high-risk women more accurately than traditional methods and determine who may need screening at an earlier age,” said Dr. Lehman. “We already screen some women between the ages of 30 and 39 when they have a clearly elevated risk due to family or genetic history. In the future, a baseline mammogram at age 30 could allow women with a high imaging-based risk score to enter that earlier, more effective screening pathway.”
In this context, the legislation on breast density currently in force in 32 states, which requires women to be informed about this parameter after a mammogram, is also relevant. For the authors of the study, this information could be expanded with new tools. “We would like women to receive information about their breast density and their risk score based on AI images,” said Dr. Lehman, who concluded: “We can do better than just look at a mammogram and say ‘it’s dense or it’s not dense’ to inform women about their risk.”
