Breast cancer does not appear suddenly. It develops over time as abnormal cells grow. What if doctors could detect these changes earlier, even before they show up on a mammogram?
A new study suggests that artificial intelligence (AI) may help. Researchers found that an AI model, when analyzing three years of mammograms, was better at identifying people at high risk of breast cancer than traditional health history questionnaires.
How AI Improves Breast Cancer Risk Prediction
Most current risk prediction tools rely on a single point in time, such as a person’s answers to a breast cancer risk questionnaire or their latest mammogram results. However, AI can use patterns from past mammograms to predict future risks more accurately.
A research team led by Dr. Shu (Joy) Jiang, a professor at Washington University, developed an AI model that analyzes up to three years of mammograms to estimate a person’s risk of developing breast cancer within five years.
The study reviewed mammograms from over 10,000 women who had screenings between 2008 and 2012, with follow-ups until 2020. Of these, 478 women were diagnosed with breast cancer. The AI model detected patterns, such as changes in breast density, to assess risk.
When tested on another 18,000 women, the AI model identified 61% of those who later developed breast cancer as high-risk. In comparison, the Tyrer-Cuzick questionnaire, a traditional risk assessment tool, only identified 26% of high-risk cases. This means the AI model was 2.3 times more accurate.
What’s Next for AI in Breast Cancer Detection?
Researchers are now testing the AI model on more diverse groups of women, including Hispanic, Pacific Islander, and Native American women. They are also exploring whether using more than three years of mammograms could improve accuracy.
Dr. Jiang and her team believe that with further research, AI could become a powerful tool for early breast cancer detection, helping doctors provide better screenings and treatments.
Source: breastcancer
