The abstract is read at the publisher; the summary is JClub's.
HEAL-Surv’s multimodal teacher-student model predicts five-year net disease-specific survival in breast cancer, reaching mean Uno’s concordance 0.723 on 423 METABRIC patients.
- Why it matters: Imaging mass cytometry and DNA methylation can be complementary for prognosis, but new patients may lack one modality, creating a need for accurate survival prediction with fewer inference inputs.
- What they did: The additive Cox teacher uses a hierarchical IMC graph encoder plus DNA methylation and clinical variables, while the student is distilled from cross-fitted teacher survival curves and observed outcomes.
- The result: Distillation improved mean five-year Uno’s concordance from 0.658 to 0.704 versus an outcome-only matched student, and an IMC-clinical variant achieved 0.660 on 265 Basel patients without refitting.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.