TL;DR
This paper presents a novel method for creating personalized 3D heart models that can adapt over time, specifically to capture cardiovascular changes during digestion. By analyzing clinical data from over 80 patients, they developed a data-driven approach to estimate and track key heart parameters, enabling more accurate and efficient simulations of physiological responses.
What Problem Does This Paper Address?
Creating accurate computational models of the heart is crucial for understanding cardiovascular diseases and testing treatments. However, these models often require a large number of parameters that are difficult to determine, especially when trying to capture dynamic physiological changes. Existing models typically focus on a single snapshot in time, making it challenging to simulate how the heart adapts to events like digestion, which involves increased blood flow and heart rate. This paper tackles the problem of longitudinally estimating these complex electromechanical parameters in 3D heart models, allowing for the simulation of dynamic cardiovascular responses over time.
Methods
The authors developed a two-step parameter estimation process. First, they calibrated both static (e.g., myocardial fiber direction) and time-varying (e.g., contractility, arterial resistance) parameters of their 3D electromechanical heart models using pre-ingestion MRI data from 21 patients. This calibration aimed to match simulated outputs like stroke volume and pressures to the measured values at this initial time point (T1).
In the second step, they reused the estimated static parameters and re-estimated the time-varying parameters to match the cardiovascular measurements at a later time point (approximately 1 hour after a meal, T2). This longitudinal approach allowed them to capture the changes in heart function during digestion.
A key innovation is the development of a "parameter evolution law." This law is derived from a multivariate regression analysis of the estimated parameter changes between T1 and T2, correlated with changes in heart rate and stroke volume. This law enables the prediction of how electromechanical parameters should change to simulate specific cardiovascular responses (like increased heart rate or stroke volume) while keeping mean and diastolic pressures constant, without needing to re-run the full, computationally expensive parameter estimation for every new simulation.
Datasets
The study utilized data from a clinical trial involving over 80 patients who underwent a high-energy meal. For the modeling, data from 21 of these patients were used. This data included:
- Short-axis cardiac cine MRI sequences acquired before and after meal ingestion.
- Measurements of stroke volume, systolic, diastolic, and mean cuff pressures at multiple time points.
- Endocardial boundary tracking from MRI sequences.
The paper does not introduce a new benchmark dataset but rather leverages existing clinical data to validate their modeling approach.
Key Insights and Ideas
- Longitudinal Modeling is Feasible: The paper demonstrates that it's possible to consistently estimate and track changes in complex 3D electromechanical heart model parameters over time using clinical data.
- Physiological Relevance of Parameter Changes: The observed trends in estimated parameters (e.g., decreased peripheral resistance, decreased arterial compliance, increased venous pressure) during digestion align with known physiological responses to increased demand, providing confidence in the model's accuracy.
- Predictive Power of the Evolution Law: The developed parameter evolution law offers a computationally efficient way to simulate dynamic cardiovascular changes, bypassing the need for repeated, intensive parameter estimation. This is a significant step towards real-time or near-real-time simulation of physiological responses.
- Inter-patient Variability is Key: The study emphasizes the importance of patient-specific modeling and acknowledges the significant inter-patient variability in how individuals respond to physiological challenges.
Key Results
- The method successfully estimated 6 parameters at T1 and 4 parameters at T2 for 21 patient-specific hearts, achieving an average fit error of 1.9 mL for Stroke Volume and 0.1 mmHg for pressures.
- On average, the peripheral resistance (R) decreased by 14.2%, arterial compliance (C) decreased, and venous pressure (Pve) increased, consistent with physiological expectations during digestion.
- The developed parameter evolution law, when tested using a leave-one-out cross-validation, accurately predicted target Stroke Volume at T2 within 1.9 mL and mean absolute variations in Diastolic and Mean Pressure within 2.1 mmHg.
Limitations and Future Work
- Limited Scope of "Digestion": The study focuses on a specific type of meal and a relatively short time window (1 hour). The authors suggest extending this to quantify how the law holds for more significant changes in cardiac output, such as those seen during more intense exercise.
- Predicting Heart Rate and Stroke Volume: While the law predicts parameter changes based on given changes in heart rate and stroke volume, future work could focus on predicting these outputs themselves and then using the law to simulate the resulting heartbeats.
- Further Validation: While the leave-one-out validation was promising, further validation with independent datasets or different physiological scenarios would strengthen the findings.
Summary updated