Latent Dynamics · 2026-10-20
CoRe-WM: Conditioned Response-State World Modeling for Longitudinal Treatment-Response Prediction in Breast DCE-MRI
Lin Zhao, Bowen Chen, Qiaohui Gao, Shaowen Wan, Xiang Li, Wei Liu, Tianming Liu
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
CoRe-WM formulates longitudinal breast DCE-MRI pCR prediction as outcome-free conditioned next-latent prediction. It learns a patient- and treatment-conditioned latent state-transition model using a JEPA-style objective, plus a factorized response state pathway supervised by image-derived response guidance. After pretraining, parameters are frozen and a linear readout predicts pCR at each decision point. Evaluated on I-SPY2 (5-fold) and transferred to I-SPY1.
Why it matters
The authors address a gap in longitudinal pCR prediction: many methods optimize multi-visit classifiers directly for the clinical endpoint, without explicitly modeling treatment-response dynamics. CoRe-WM shifts learning to a state-transition/world-model objective and then freezes representations for pCR readout, enabling progressive risk updating from observed prefixes while incorporating patient profile and assigned therapy into the transition dynamics.
Method
- Outcome-free pretraining: learn a conditioned latent state-transition model to predict the next latent patient state from an observed state prefix and condition c=(baseline profile, treatment), using a conditioned causal Transformer within JEPA.
- Factorized response-state pathway: a parallel low-dimensional response state encodes lesion morphology/geometry priors, guided by an image-derived response guidance (IRG) loss, then used for latent correction of the next-state prediction.
- After pretraining, freeze all parameters and use a frozen linear readout over accumulated response states to predict binary pCR at each clinical decision point.
Limitation
The authors state that improving image-only representation learning under limited longitudinal data is an important direction for future work, because the explicit geometry pathway/response guidance partially compensate but the model relies on handcrafted geometry descriptors and capacity to discover novel imaging biomarkers is limited. They also note that independent external validation on larger/more diverse cohorts is limited by availability of longitudinal DCE-MRI datasets with matched clinical data and annotations.
Abstract (from arXiv)
Pathologic complete response (pCR) after neoadjuvant chemotherapy is a key prognostic indicator in aggressive breast cancer, yet existing prediction approaches either rely on pretreatment imaging alone, missing the patient's evolving response, or optimize multi-visit classifiers under direct outcome supervision, coupling representation learning with the clinical endpoint. We formulate longitudinal pCR prediction as a world-modeling problem and propose Conditioned Response-state World Modeling (CoRe-WM), which learns a latent state-transition model that predicts the next imaging state from the observed sequence. The transition function is conditioned on patient clinical profile and assigned therapy through additive shift and feature-wise modulation, enabling patient- and treatment-specific dynamics. A Factorized Response State pathway further decouples treatment-response dynamics from imaging appearance by encoding the structural prior that response manifests primarily in lesion morphology. After pretraining, all parameters are frozen and a linear readout predicts pCR at each clinical decision point. We evaluate CoRe-WM on I-SPY2 under five-fold cross-validation and assess cross-cohort transfer on I-SPY1. CoRe-WM achieves the highest discrimination among all evaluated methods at every prospective decision point, and demonstrates robust cross-cohort transfer. Ablation studies identify patient-profile conditioning and the factorized response pathway as the principal contributors. These results establish world-model-based representation learning as a promising framework for outcome prediction in oncology and beyond.
Related papers
- CaMeWM: A Causal Medical World Model for Generative Tumor Evolution and Treatment Planning
- Patient, Place, Prior (P$^3$): What Counts as Personalization in Medical World Models?
- PatientWM: Towards Generalizable Clinical World Models via Unified Patient States
- Forecasting Patient Trajectories from the Electronic Health Records with Action-Conditioned World Models
- Repeated-Measure Leakage, Distribution Shift, and Reliability under Partial Observation in Patient World Models
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-18 · Indexed by Pub-AI: 2026-10-20 UTC
CoRe-WM: Conditioned Response-State World Modeling for Longitudinal Treatment-Response Prediction in Breast DCE-MRI
Preprint posted: 2026-10-18 · Original record ↗
- arxiv
- 2610.25333
Explicit source versions (1)
- v1 · 2026-10-18
- Preprint available
Source date: 2026-10-18 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
Pathologic complete response (pCR) after neoadjuvant chemotherapy is a key prognostic indicator in aggressive breast cancer, yet existing prediction approaches either rely on pretreatment imaging alone, missing the patient's evolving response, or optimize multi-visit classifiers under direct outcome supervision, coupling representation learning with the clinical endpoint. We formulate longitudinal pCR prediction as a world-modeling problem and propose Conditioned Response-state World Modeling (CoRe-WM), which learns a latent state-transition model that predicts the next imaging state from the observed sequence. The transition function is conditioned on patient clinical profile and assigned therapy through additive shift and feature-wise modulation, enabling patient- and treatment-specific dynamics. A Factorized Response State pathway further decouples treatment-response dynamics from imaging appearance by encoding the structural prior that response manifests primarily in lesion morphology. After pretraining, all parameters are frozen and a linear readout predicts pCR at each clinical decision point. We evaluate CoRe-WM on I-SPY2 under five-fold cross-validation and assess cross-cohort transfer on I-SPY1. CoRe-WM achieves the highest discrimination among all evaluated methods at every prospective decision point, and demonstrates robust cross-cohort transfer. Ablation studies identify patient-profile conditioning and the factorized response pathway as the principal contributors. These results establish world-model-based representation learning as a promising framework for outcome prediction in oncology and beyond.