Researchers from the University of Colorado Boulder, UC Anschutz Medical Campus, Georgia Institute of Technology, and Emory University developed a deep learning AI model to predict breast cancer risk by analyzing how single cells respond to pressure. The AI system, detailed in *Nature Biomedical Engineering* in May 2024, learned to identify patterns in the biomechanical properties of breast cells, offering a novel method for early risk assessment. This work was led by researchers Bo Guo, Randy G. Seifert, and senior author Lisa A. Forrest.
The team utilized images of single cells taken from diagnostic mammograms of 80 women – 40 previously diagnosed with breast cancer and 40 without. By observing how these cells deformed under pressure, the AI model was trained to detect subtle mechanical differences indicative of disease progression. Bo Guo stated, 'Our AI model represents a completely new way to predict breast cancer risk by quantifying changes in tissue stiffness and cell mechanics at the earliest stages of the disease.'
The developed AI model demonstrated significant accuracy, identifying 92.5% of women who developed cancer within five years and 97.5% of those who did not. This non-invasive approach could complement existing screening methods like mammograms, potentially improving personalized risk prediction and early detection strategies for breast cancer.
