This paper presents a novel framework integrating Bayesian physiological modeling with optimal control strategies to achieve uncertainty-aware, personalized hemodynamic regulation during fluid resuscitation. An uncertainty-aware variational autoencoder state-space model (UVAE-SSM) was first developed to capture the dynamical relationship between mean arterial pressure (MAP) and fluid infusion using limited data, while explicitly modeling aleatoric uncertainty (i.e., randomness in the measurements, such as sensor noise). Then, a Bayesian nonlinear state-space model (BNSSM) was developed by utilizing Bayesian neural networks (BNNs) to capture epistemic uncertainty arising from physiological and patient-specific variability, enabling the creation of a virtual patient generator (VPG). Building on this uncertainty-aware modeling framework, a stochastic radial basis function model predictive control (sRBF-MPC) algorithm was designed to track the MAP target while satisfying physiological constraints. Finally, an online fine-tuning algorithm was developed to adapt the nominal UVAE-SSM using streaming VPG data, enabling progressive personalization during closed-loop therapy. Simulation results across unseen animal subjects and an independent human clinical dataset demonstrated the strong predictive accuracy and cross-population generalizability of the UVAE-SSM and BNSSM models. Closed-loop evaluations confirmed that the proposed sRBF-MPC framework achieved stable MAP regulation while providing better risk-aware control compared to quadratic MPC (Q-MPC) and stochastic quadratic MPC (sQ-MPC). Overall, the proposed framework accounts for inter- and intra-patient variability through online model adaptation, offering a promising step toward uncertainty-aware, personalized hemodynamic modeling and control in critical care.
Figures & tables
Fig. 1: Proposed uncertainty-aware modeling framework. (a) UVAE-SSM as nominal model: The encoder maps a sliding window of NA MAP measurements and infusion rates (yk−NA:k,uk−NA:k) to the parameters (μx,σx) of a Gaussian latent distribution. A latent state xk is drawn via the reparameterization trick and passed to the deterministic transition network, which predicts the next latent state x^k+1 given (xk,uk) . The probabilistic decoder maps the latent state to a predictive MAP distribution N(μy,σy2) , explicitly capturing aleatoric uncertainty in the hemodynamic output. Dashed lines indicate training-only connections used to provide supervision targets for the transition loss. (b) BNSSM as VPG : The encoder and decoder form a deterministic autoencoder that maps observations to latent states and reconstructs MAP. The transition function fβ(xk,uk) is implemented as a BNN, where weights β are treated as random variables with variational posterior qη(β) , approximated via Bayes by Backprop. At inference, β is sampled once per complete simulation trajectory from qη(β) ; each distinct draw yields one virtual patient realization.
Fig. 2: Structure of the closed-loop fluid resuscitation framework with online personalization.
Fig. 3: MAP prediction for held-out animal test subjects using the proposed UVAE-SSM and the lumped hemodynamic model [ 6 ] .
Fig. 4: MAP prediction for human subjects using UVAE-SSM.
UVAE-SSM (Animal)
Lumped Model (Animal)
UVAE-SSM (Human)
Metric
Subject1
Subject2
Subject1
Subject2
Subject1
Subject2
Subject3
Subject4
Subject5
Subject6
RMSE (%)
4.66
3.07
7.74
7.19
2.49
0.86
0.91
2.71
0.54
0.53
MAE (%)
2.10
1.88
5.84
5.38
1.76
0.59
0.44
1.14
0.38
0.38
MDAPE (%)
0.65
1.02
5.57
3.81
1.13
0.39
0.23
0.49
0.25
0.26
TABLE I: MAP prediction performance of the UVAE-SSM on held-out animal and human test subjects. The lumped hemodynamic model [ 6 ] is included as a physiologically grounded baseline for the animal subject evaluation.
Fig. 5: MAP prediction on held-out animal test subjects using the VPG. The dashed line shows the ensemble mean over N=50 virtual patients.
Fig. 6: MAP prediction on held-out human clinical test subjects using the VPG.
VPG (Animal)
VPG (Human)
Metric
Subject1
Subject2
Subject1
Subject2
Subject3
Subject4
Subject5
Subject6
RMSE (%)
6.91
3.35
2.38
0.87
0.54
0.63
0.42
0.57
MAE (%)
4.01
2.04
2.04
0.60
0.43
0.42
0.31
0.42
MDAPE (%)
1.84
1.13
1.71
0.32
0.40
0.20
0.21
0.33
TABLE II: MAP prediction performance of the VPG on held-out animal and human test subjects
Fig. 7: Closed-loop MAP regulation under four hemorrhage scenarios comparing Q-MPC, sQ-MPC, and sRBF-MPC. Each panel shows the patient MAP trajectory (top) and the corresponding infusion rate (bottom). The light-red bands mark the active hemorrhage window ( t=0 – 15 min); The light-yellow region marks the pre-control phase, during which no infusion is applied. The dotted vertical line at t=30 min marks controller activation. Dashed purple vertical lines at t=60,90,120, and 150 min indicate periodic fine-tuning events. The dashed gray and red horizontal lines indicate the MAP target (80 mmHg) and the safety floor (60 mmHg), respectively.
Scenario
Controller
RMSE (%)
MAE (%)
MDAPE (%)
Below 60 mmHg (%)
Total Infusion (mL)
Solve Time/iteration (ms)
1
Q-MPC
11.09
7.61
4.58
2.9
3100.8
27.5±17.3
sQ-MPC
10.67
7.69
4.52
2.8
3011.8
62.3±65.7
sRBF-MPC
8.13
5.67
2.57
0.6
3364.4
255.9±326.0
2
Q-MPC
3.67
3.44
3.40
0.0
2145.8
14.0±6.9
sQ-MPC
3.68
3.45
3.40
0.0
2143.5
28.8±15.2
sRBF-MPC
2.20
2.04
2.10
0.0
2663.8
55.4±31.3
TABLE III: Closed-Loop Performance Comparison of Q-MPC, sQ-MPC, and sRBF-MPC Across Four Hemorrhage Scenarios
Electrocardiogram (ECG)-based models have achieved strong performance in diagnostic tasks, yet they remain limited in modeling how cardiac dynamics evolve under external interventions. In particular, existing approaches focus primarily on static prediction and lack mechanisms to capture ECG variations under different pharmacological conditions. In this work, we propose an ECG World Model for action-conditioned predictive simulation of cardiac electrophysiology. Moving beyond disjoint pipelines, our framework features a principled integration of physiological ordinary differential equation (ODE) priors into latent diffusion dynamics via energy regularization. This structural constraint enables the synthesis of physiologically plausible post-intervention ECG trajectories while effectively mitigating generative hallucinations. Building on this simulation process, we introduce an uncertainty-aware evaluation strategy that leverages the stochasticity of diffusion sampling to characterize both the expected clinical risk and its variability, allowing a more reliable comparative assessment of candidate interventions. We evaluate our method across diverse settings, including controlled drug-response scenarios and real-world clinical records. Beyond standard waveform metrics, experimental results demonstrate improved risk calibration and strong alignment with expert-informed treatment preferences. These results establish our approach as a robust foundation for safe and intervention-aware clinical decision support.
Zhikang Chen, Yue Wang, Sen Cui +4
University of Oxford · Tsinghua University · Southern University of Science and Technology
Respiratory support can shape the short-term physiological trajectory of critically ill patients, but patients receiving the same intervention may follow different physiological trajectories. Clinical patient dynamics models typically predict future states from recent physiology and recorded interventions, while physiological change is mainly represented through the predicted future state. We propose a response-aware patient dynamics model that explicitly represents physiological change during autoregressive state updating. The model decomposes predicted physiological change into state-dependent baseline dynamics and respiratory-support-associated deviations, with room air providing a reference for the decomposition. We provide a formal analysis of this reference-anchored formulation. A response pathway encodes the predicted physiological change and uses it to update the latent patient state across the forecast horizon. Across ICU cohorts from two independent institutions, the proposed model achieves comparable overall trajectory prediction to patient dynamics baselines, with more consistent improvements when physiological states are changing.
Owing to the recent advancements in wearable devices for health care, the importance of BP estimation without cuffs increases. Cuff technologies are inappropriate for continuous BP measurement due to their inconvenient usage, invasive character, necessity of calibration, large size, and inability to perform long-term monitoring. Normally, the algorithm used for cuffless BP prediction employs machine learning models that operate according to the data-driven approach. However, although they show high numerical accuracy, ML models do not provide any interpretability, resulting in poor physiological validity and clinical applicability. We propose a combination of Windkessel and ML models that incorporates the physical aspects into the latter one. It is performed by reformulating Windkessel into a form that will allow employing ML models. The result is a system of ODEs which can be used in the neural network. Thus, the inclusion of physical constraints improves the data-driven approach by making models consistent with physics, understandable, and robust. For illustration, we apply the described technique using a publicly available MIMIC-II database that we obtain from the UCI Machine Learning Repository.
Vaibhav Gollapalli, Aniruth Ananthanarayanan
Department of Physics University of North Texas Denton, United States of America