Changes in Breast Cancer Risk Scores
Researchers published findings in Radiology detailing how artificial intelligence can track changes in breast cancer risk scores over time using screening mammograms, a development that could transform risk assessment for the disease. The study, led by Dr. Constance D. Lehman of Harvard Medical School, analyzed 158,807 mammograms from 54,014 women between 2009 and 2019, revealing that AI-generated risk scores for cancer patients increased progressively over six years, while scores for cancer-free individuals remained stable.
Traditional risk models, which rely on factors like family history or genetic mutations, have limited effectiveness in population-based screening, according to the study. However, the deep learning model used in this research evaluated entire mammogram images rather than predefined features like density, achieving better accuracy in predicting five-year risk. Among the 817 women diagnosed with breast cancer within a year of their final mammogram, AI risk scores rose from a median of 2.1 to 6.6 in the six years prior to diagnosis.
Why does this matter?
The findings highlight a critical gap in current breast cancer screening. Most women diagnosed with the disease lack known genetic risk factors, making traditional models less effective. The AI system’s ability to detect subtle, evolving patterns in mammograms could improve early detection for the majority of cases, which are sporadic rather than inherited. Dr. Lehman emphasized that 85% of breast cancer cases occur without significant family history, underscoring the potential of this technology to address unmet needs in risk assessment.
What may happen next?
Further validation studies could lead to the integration of AI risk scores into routine screening protocols. If adopted, this approach might reduce unnecessary biopsies or increase monitoring for high-risk patients. However, implementation would require regulatory approval and training for radiologists to interpret dynamic risk scores. Researchers also note the need to ensure the model’s performance across diverse populations and imaging settings.
Experts suggest that the study’s methodology—using only imaging data without demographic or clinical information—could simplify risk assessment while reducing bias. However, challenges remain in translating these results into clinical practice, including technical hurdles and ethical considerations around data privacy.
As AI continues to reshape medical diagnostics, this research marks a step toward more nuanced, data-driven approaches to breast cancer prevention. However, its real-world impact will hinge on how effectively healthcare systems adapt to these evolving tools.
What is the key finding of the study? AI models can detect changes in breast cancer risk scores over time using mammograms, with scores rising significantly in women who later develop cancer.
How does this differ from traditional risk models? Unlike models relying on family history or genetics, this AI system analyzes entire mammogram images, identifying patterns invisible to the human eye.
What are the next steps for this research? Further validation studies are needed to confirm the model’s effectiveness across diverse populations and to explore its integration into clinical guidelines.
Could this technology redefine how breast cancer risk is assessed in the coming years?