Pilot use-case – research demonstrator
Interval breast cancer risk at 12 and 24 months (from 2–3 successive mammograms)
Longitudinal rare-event modeling with censoring, weak imaging signals, and uncertainty bands.
Research-only disclaimer
These demonstrator use cases are provided for scientific research and exploration purposes only.
They do not constitute medical diagnosis, clinical decision support, or treatment recommendation.
Ethical and Regulatory Note
Read the cross-cutting note →Research question
Can AI-assisted longitudinal analysis of successive mammograms help identify weak imaging signals associated with the occurrence of interval breast cancer within 12 and 24 months, in a research setting?
Explicit exclusions (non-negotiable)
- No cancer detection
- No screening replacement
- No clinical alerting
Data structure (illustrative)
- 2 or 3 temporally ordered mammograms
- Imaging-derived features (abstracted)
- Time-to-event structure
- Censoring explicitly modeled
Modeling concepts (demo-level)
- Longitudinal feature evolution
- Survival / time-to-event modeling
- Relative risk trajectories with uncertainty bands
- Sensitivity to missing time points