Latent Dynamics · 2026-10-30
PulmoWorld: An Asynchronous Multimodal Medical World Model for Longitudinal Pneumonia Dynamics
Cheng Tang, Xinyu Zhang, Jiawei Chen, Minghui Zhang, Shaocong Wei, Wei Li, Yuxing Chen, Kongcheng Zhang, Chun-Mei Feng, Nanqing Dong, Xinyuan Zhu, Ziyan Huang, Junjun He, Yirong Chen, Lingtong Huang, Min Cen, Ming Hu
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
PulmoWorld is an asynchronous multimodal medical world model for longitudinal pneumonia dynamics. It uses history-aware filtering to propagate a shared patient belief forward in time and apply observation-driven corrections only when measurements are observed. The resulting belief supports recursive simulation for forecasts and evidence-constrained planning via Cap-MBRL, which searches within an authorized action set and uses risk-budgeted replacement of a frozen fallback.
Why it matters
The authors address irregular, incomplete multimodal clinical evidence by separating state propagation from observation correction, reducing issues caused by fusing measurements of different ages. They also connect the learned asynchronous belief to recursive forecasting and evidence-constrained action selection, and evaluate prognostic and planning performance on pneumonia cohorts and a known-mechanism simulator.
Method
- History-aware filtering: propagate a patient belief with time elapsed, then apply a masked innovation correction only for observed measurement groups; missing measurements produce no correction.
- Shared-state recursive dynamics: decode future clinical values and modality representations from recursively propagated latent states, with care-conditioned branching for simulated trajectories.
- Evidence-constrained planning (Cap-MBRL): compare candidate continuations within an authorized action set using model rollouts; replace a frozen fallback only if the candidate meets a risk budget and improves predicted cost by a margin δ.
Limitation
The authors state that MIMIC-IV and MIMIC-CXR records must be obtained under the PhysioNet data-use agreement and are not redistributed; MASS patient records were used with permission after personal identifying information was removed and are not part of the public release.
Abstract (from arXiv)
Pneumonia changes across respiration, inflammation, and organ function, while the measurements that describe it arrive at different times and are often incomplete. We introduce PULMOWORLD, a world model for longitudinal pneumonia dynamics that connects asynchronous state estimation, recursive simulation, and evidence-constrained planning. History-aware filtering separates state propagation from observation-driven correction, so a measurement updates the shared belief only when it is observed. That belief supports recursive forecasts of clinical values and modality representations, and summaries of future prognosis. Cap-MBRL searches imagined continuations inside an authorized action set and replaces a frozen fallback only when a candidate meets the risk budget and a predicted-cost margin. On the pneumonia cohort MASS, recursive forecasts are competitive with task-trained sequence models, and simulated future states retain prognostic information. Trained separately on public MIMIC-IV-CXR chest radiographs and clinical trajectories, this model improves prognostic discrimination over latest-value fusion in development and is competitive on long-horizon error on the locked pneumonia panel. In five known-mechanism MASS-Syn environments, Cap-MBRL lowers selection regret relative to direct fallback execution.
Related papers
- 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
- Understanding Trajectory Heterogeneity in Federated World Model Learning
- Patient, Place, Prior (P$^3$): What Counts as Personalization in Medical World Models?
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First announced by arXiv: 2026-10-30 · Indexed by Pub-AI: 2026-10-30 UTC
PulmoWorld: An Asynchronous Multimodal Medical World Model for Longitudinal Pneumonia Dynamics
First announced by arXiv: 2026-10-30 · Original record ↗
- arxiv
- 2610.38289
Explicit source versions (1)
- v1 · Date not supplied
- Preprint available
Source date: 2026-10-30 · First detected: 2026-10-30 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · RSS guid=oai:arXiv.org:2610.38289v1; arxiv:announce_type=new; item pubDate (official metadata)
Abstract · arxiv
Pneumonia changes across respiration, inflammation, and organ function, while the measurements that describe it arrive at different times and are often incomplete. We introduce PULMOWORLD, a world model for longitudinal pneumonia dynamics that connects asynchronous state estimation, recursive simulation, and evidence-constrained planning. History-aware filtering separates state propagation from observation-driven correction, so a measurement updates the shared belief only when it is observed. That belief supports recursive forecasts of clinical values and modality representations, and summaries of future prognosis. Cap-MBRL searches imagined continuations inside an authorized action set and replaces a frozen fallback only when a candidate meets the risk budget and a predicted-cost margin. On the pneumonia cohort MASS, recursive forecasts are competitive with task-trained sequence models, and simulated future states retain prognostic information. Trained separately on public MIMIC-IV-CXR chest radiographs and clinical trajectories, this model improves prognostic discrimination over latest-value fusion in development and is competitive on long-horizon error on the locked pneumonia panel. In five known-mechanism MASS-Syn environments, Cap-MBRL lowers selection regret relative to direct fallback execution.