AI Tracks Mammogram Changes to Forecast Breast Cancer Risk
A new deep‑learning model can predict a woman’s future breast cancer risk by analysing how her mammogram images evolve over time. The approach could transform routine screening into a personalised risk‑assessment tool, helping clinicians spot high‑risk patients earlier.
A breakthrough deep‑learning system can now read subtle changes in a woman’s mammograms and translate them into a future breast‑cancer risk score. By comparing images taken at different times, the model identifies patterns that precede tumour development, offering a new way to personalise screening schedules.
What Changed?
Researchers trained a neural network on thousands of paired mammograms from women who later developed breast cancer and those who did not. The algorithm learns to quantify how the breast tissue’s appearance shifts over months and years, then uses those shifts to estimate the probability of cancer within the next few years.
Key Findings
- Tracking image changes outperforms static risk models that rely only on a single scan.
- The AI’s risk scores correlate strongly with established clinical risk factors, yet add independent predictive power.
- In validation studies, the model correctly identified 80% of women who developed cancer, with a false‑positive rate below 10%.
- When combined with genetic risk scores, the predictive accuracy improves further, suggesting a multimodal approach.
Why It Matters
Current screening guidelines recommend fixed intervals for mammography, often missing the window when a tumour is most treatable. A dynamic, image‑based risk score could:
- Tailor screening frequency to each patient’s evolving risk.
- Reduce unnecessary imaging and associated anxiety.
- Enable earlier intervention for those at high risk, potentially improving survival rates.
Expert View
Dr. Kefah Mokbel, a leading breast‑cancer researcher, notes that “personalised screening is the future.” She highlights that AI can serve as a triage partner, flagging patients who need more intensive follow‑up while reassuring low‑risk individuals.
Context
Artificial intelligence has already shown promise in image interpretation, but this is the first study to demonstrate that longitudinal changes in mammograms carry actionable risk information. The work builds on earlier reports that AI can detect early signs of cancer and now extends that capability to risk prediction.
What to Watch Next
Clinical trials are underway to test the algorithm in real‑world screening programs. If successful, health authorities may incorporate AI‑derived risk scores into national guidelines, moving from a one‑size‑fits‑all approach to a truly personalised model of care.
Discover more from NewzQuest
Subscribe to get the latest posts sent to your email.
