Critical Care

Cross-Domain Bridges

Bridge Neural controlled differential equations bridge rough-path theory and irregular ICU trajectory modeling for event forecasting under missingness.

Fields: Critical Care, Machine Learning, Stochastic Processes

Speculative analogy (to be empirically validated): neural CDEs translate irregularly sampled physiologic streams into continuous control paths, mirroring how rough-path summaries preserve temporal sig...

Bridge Delay-embedding reconstructions can transfer from nonlinear dynamics to ICU deterioration early-warning indicators.

Fields: Dynamical Systems, Critical Care, Signal Processing

Speculative analogy: Delay-embedding reconstructions can transfer from nonlinear dynamics to ICU deterioration early-warning indicators....

Bridge Long short-term memory dynamics connect sequence-learning memory gates with ICU physiology forecasting.

Fields: Computer Science, Critical Care, Physiology

Speculative analogy: LSTM gating provides a sequence-memory abstraction that can capture delayed physiological interactions in ICU time-series forecasting....

Open Unknowns (3)

Unknown What failure boundaries determine when `b-delay-embedding-x-icu-deterioration-early-warning` remains decision-useful? u-embedding-dimension-selection-for-icu-trajectory-instability-detection

Unknown Which missingness models are required for safe deployment of `b-lstm-sequence-memory-x-icu-physiology-forecasting`? u-missingness-aware-lstm-training-for-icu-forecasts

Unknown How robust are neural CDE ICU forecasts to clinically realistic, nonrandom missingness patterns? u-neural-cde-icu-robustness-to-missingness-patterns

Active Hypotheses

Hypothesis Methods transferred from `b-delay-embedding-x-icu-deterioration-early-warning` improve target outcomes versus domain-specific baselines at matched cost. high

Hypothesis Explicit missingness encoding in LSTM models improves ICU decompensation horizon accuracy versus simple imputation baselines. high

Hypothesis Neural CDE models improve clinically usable ICU event lead-time at fixed false-alert rate compared with interpolation-based baselines. high

Hypothesis Sepsis comprises at least two reproducible genomic endotypes (SRS1: immunosuppressed, higher mortality; SRS2: immunoactivated, lower mortality) identifiable from whole- blood transcriptomics within 24h of admission, with differential response to corticosteroids: steroids harm SRS1 patients by further suppressing immunity and benefit SRS2 by reducing hyperinflammation. high

Know something about Critical Care? Contribute an unknown or hypothesis →

Generated 2026-05-10 · USDR Dashboard