Model-based identification and diagnosis of a porcine model of induced endotoxic shock with hemofiltration

Preserving multi-dimensional information: A hypersphere method for parameter space analysis

Background

Physiological modelling often involves models described by large numbers of variables and significant volumes of clinical data. Mathematical interpretation of such models frequently necessitates analysing data points in high-dimensional spaces. Existing algorithms for analysing high-dimensional points either lose important dimensionality or do not describe the full position of points. Hence, there is a need for an algorithm which preserves this information.

Methods

The most-distant uncovered point (MDUP) hypersphere method is a binary classification approach which defines a collection of equidistant N-dimensional points as the union of hyperspheres. The method iteratively generates hyperspheres at the most distant point in the interest region not yet contained within any hypersphere, until the entire region of interest is defined by the union of all generated hyperspheres. This method is tested on a 7-dimensional space with up to 35.8 million points representing feasible and infeasible spaces of model parameters for a clinically validated cardiovascular system model.

Results

For different numbers of input points, the MDUP hypersphere method tends to generate large spheres away from the boundary of feasible and infeasible points, but generates the greatest number of relatively much smaller spheres around the boundary of the region of interest to fill this space. Runtime scales quadratically, in part because the current MDUP implementation is not parallelised.

Conclusions

The MDUP hypersphere method can define points in a space of any dimension using only a collection of centre points and associated radii, making the results easily interpretable. It can identify large continuous regions, and in many cases capture the general structure of a region in only a relative few hyperspheres. The MDUP method also shows promise for initialising optimisation algorithm starting conditions within pre-defined feasible regions of model parameter spaces, which could improve model identifiability and the quality of optimisation results.

Blood pressure waveform contour analysis for assessing peripheral resistance changes in sepsis

BACKGROUND

This paper proposes a methodology for helping bridge the gap between the complex waveform information frequently available in an intensive care unit and the simple, lumped values favoured for rapid clinical diagnosis and management. This methodology employs a simple waveform contour analysis approach to compare aortic, femoral and central venous pressure waveforms on a beat-by-beat basis and extract lumped metrics pertaining to the pressure drop and pressure-pulse amplitude attenuation as blood passes through the various sections of systemic circulation.

RESULTS

Validation encompasses a comparison between novel metrics and well-known, analogous clinical metrics such as mean arterial and venous pressures, across an animal model of induced sepsis. The novel metric Ofe → vc, the direct pressure offset between the femoral artery and vena cava, and the clinical metric, ΔMP, the difference between mean arterial and venous pressure, performed well. However, Ofe → vc reduced the optimal average time to sepsis detection after endotoxin infusion from 46.2 min for ΔMP to 11.6 min, for a slight increase in false positive rate from 1.8 to 6.2%. Thus, the novel Ofe → vc provided the best combination of specificity and sensitivity, assuming an equal weighting to both, of the metrics assessed.

CONCLUSIONS

Overall, the potential of these novel metrics in the detection of diagnostic shifts in physiological behaviour, here driven by sepsis, is demonstrated.

Next-generation, personalised, model-based critical care medicine: a state-of-the art review of in silico virtual patient models, methods, and cohorts, and how to validation them

Critical care, like many healthcare areas, is under a dual assault from significantly increasing demographic and economic pressures. Intensive care unit (ICU) patients are highly variable in response to treatment, and increasingly aging populations mean ICUs are under increasing demand and their cohorts are increasingly ill. Equally, patient expectations are growing, while the economic ability to deliver care to all is declining. Better, more productive care is thus the big challenge. One means to that end is personalised care designed to manage the significant inter- and intra-patient variability that makes the ICU patient difficult. Thus, moving from current "one size fits all" protocolised care to adaptive, model-based "one method fits all" personalised care could deliver the required step change in the quality, and simultaneously the productivity and cost, of care. Computer models of human physiology are a unique tool to personalise care, as they can couple clinical data with mathematical methods to create subject-specific models and virtual patients to design new, personalised and more optimal protocols, as well as to guide care in real-time. They rely on identifying time varying patient-specific parameters in the model that capture inter- and intra-patient variability, the difference between patients and the evolution of patient condition. Properly validated, virtual patients represent the real patients, and can be used in silico to test different protocols or interventions, or in real-time to guide care. Hence, the underlying models and methods create the foundation for next generation care, as well as a tool for safely and rapidly developing personalised treatment protocols over large virtual cohorts using virtual trials. This review examines the models and methods used to create virtual patients. Specifically, it presents the models types and structures used and the data required. It then covers how to validate the resulting virtual patients and trials, and how these virtual trials can help design and optimise clinical trial. Links between these models and higher order, more complex physiome models are also discussed. In each section, it explores the progress reported up to date, especially on core ICU therapies in glycemic, circulatory and mechanical ventilation management, where high cost and frequency of occurrence provide a significant opportunity for model-based methods to have measurable clinical and economic impact. The outcomes are readily generalised to other areas of medical care.

Minimally invasive, patient specific, beat-by-beat estimation of left ventricular time varying elastance

Background

The aim of this paper was to establish a minimally invasive method for deriving the left ventricular time varying elastance (TVE) curve beat-by-beat, the monitoring of which’s inter-beat evolution could add significant new data and insight to improve diagnosis and treatment. The method developed uses the clinically available inputs of aortic pressure, heart rate and baseline end-systolic volume (via echocardiography) to determine the outputs of left ventricular pressure, volume and dead space volume, and thus the TVE curve. This approach avoids directly assuming the shape of the TVE curve, allowing more effective capture of intra- and inter-patient variability.

Results

The resulting TVE curve was experimentally validated against the TVE curve as derived from experimentally measured left ventricular pressure and volume in animal models, a data set encompassing 46,318 heartbeats across 5 Piétrain pigs. This simulated TVE curve was able to effectively approximate the measured TVE curve, with an overall median absolute error of 11.4% and overall median signed error of −2.5%.

Conclusions

The use of clinically available inputs means there is potential for real-time implementation of the method at the patient bedside. Thus the method could be used to provide additional, patient specific information on intra- and inter-beat variation in heart function.

Electronic supplementary material

The online version of this article (doi:10.1186/s12938-017-0338-7) contains supplementary material, which is available to authorized users.

Simulation of left atrial function using a multi-scale model of the cardiovascular system

During a full cardiac cycle, the left atrium successively behaves as a reservoir, a conduit and a pump. This complex behavior makes it unrealistic to apply the time-varying elastance theory to characterize the left atrium, first, because this theory has known limitations, and second, because it is still uncertain whether the load independence hypothesis holds. In this study, we aim to bypass this uncertainty by relying on another kind of mathematical model of the cardiac chambers. In the present work, we describe both the left atrium and the left ventricle with a multi-scale model. The multi-scale property of this model comes from the fact that pressure inside a cardiac chamber is derived from a model of the sarcomere behavior. Macroscopic model parameters are identified from reference dog hemodynamic data. The multi-scale model of the cardiovascular system including the left atrium is then simulated to show that the physiological roles of the left atrium are correctly reproduced. This include a biphasic pressure wave and an eight-shaped pressure-volume loop. We also test the validity of our model in non basal conditions by reproducing a preload reduction experiment by inferior vena cava occlusion with the model. We compute the variation of eight indices before and after this experiment and obtain the same variation as experimentally observed for seven out of the eight indices. In summary, the multi-scale mathematical model presented in this work is able to correctly account for the three roles of the left atrium and also exhibits a realistic left atrial pressure-volume loop. Furthermore, the model has been previously presented and validated for the left ventricle. This makes it a proper alternative to the time-varying elastance theory if the focus is set on precisely representing the left atrial and left ventricular behaviors.

Evaluation of a model-based hemodynamic monitoring method in a porcine study of septic shock

INTRODUCTION

The accuracy and clinical applicability of an improved model-based system for tracking hemodynamic changes is assessed in an animal study on septic shock.

METHODS

This study used cardiovascular measurements recorded during a porcine trial studying the efficacy of large-pore hemofiltration for treating septic shock. Four Pietrain pigs were instrumented and induced with septic shock. A subset of the measured data, representing clinically available measurements, was used to identify subject-specific cardiovascular models. These models were then validated against the remaining measurements.

RESULTS

The system accurately matched independent measures of left and right ventricle end diastolic volumes and maximum left and right ventricular pressures to percentage errors less than 20% (except for the 95th percentile error in maximum right ventricular pressure) and all R(2) > 0.76. An average decrease of 42% in systemic resistance, a main cardiovascular consequence of septic shock, was observed 120 minutes after the infusion of the endotoxin, consistent with experimentally measured trends. Moreover, modelled temporal trends in right ventricular end systolic elastance and afterload tracked changes in corresponding experimentally derived metrics.

CONCLUSIONS

These results demonstrate that this model-based method can monitor disease-dependent changes in preload, afterload, and contractility in porcine study of septic shock.

Validation of subject-specific cardiovascular system models from porcine measurements

A previously validated mathematical model of the cardiovascular system (CVS) is made subject-specific using an iterative, proportional gain-based identification method. Prior works utilised a complete set of experimentally measured data that is not clinically typical or applicable. In this paper, parameters are identified using proportional gain-based control and a minimal, clinically available set of measurements. The new method makes use of several intermediary steps through identification of smaller compartmental models of CVS to reduce the number of parameters identified simultaneously and increase the convergence stability of the method. This new, clinically relevant, minimal measurement approach is validated using a porcine model of acute pulmonary embolism (APE). Trials were performed on five pigs, each inserted with three autologous blood clots of decreasing size over a period of four to five hours. All experiments were reviewed and approved by the Ethics Committee of the Medical Faculty at the University of Liege, Belgium. Continuous aortic and pulmonary artery pressures (P(ao), P(pa)) were measured along with left and right ventricle pressure and volume waveforms. Subject-specific CVS models were identified from global end diastolic volume (GEDV), stroke volume (SV), P(ao), and P(pa) measurements, with the mean volumes and maximum pressures of the left and right ventricles used to verify the accuracy of the fitted models. The inputs (GEDV, SV, P(ao), P(pa)) used in the identification process were matched by the CVS model to errors <0.5%. Prediction of the mean ventricular volumes and maximum ventricular pressures not used to fit the model compared experimental measurements to median absolute errors of 4.3% and 4.4%, which are equivalent to the measurement errors of currently used monitoring devices in the ICU (∼5-10%). These results validate the potential for implementing this approach in the intensive care unit.

Assessment of ventricular contractility and ventricular-arterial coupling with a model-based sensor

Estimation of ventricular contractility and ventricular arterial coupling is clinically important in diagnosing and treating cardiac dysfunction in the critically ill. However, experimental assessment of indexes of ventricular contractility, such as the end-systolic pressure-volume relationship, requires a highly invasive maneuver and measurements that are not typical in an intensive care unit (ICU). This research describes the use of a previously validated cardiovascular system model and parameter identification process to evaluate the right ventricular arterial coupling in septic shock. Model-based ventricular arterial coupling is defined by the ratio of the end systolic right ventricular elastance (E(esrvf)) over the pulmonary artery elastance (E(pa)) or the mean pulmonary inflow resistance (R(pulin)). Results are compared to the clinical gold-standard assessment (conductance catheter method). Six anesthetized healthy pigs weighing 20-30kg received a 0.5mg kg(-1) endotoxin infusion over a period of 30min from T0 to T30, to induce septic shock and veno-venous hemofiltration was used from T60 onward. The results show good agreement with the gold-standard experimental assessment. In particular, the model-based right ventricular elastance (E(esrvf)) correlates well with the clinical gold standard (R(2)=0.69) and the model-based non-invasive coupling (E(esrvf)/R(pulin)) follow the same trends and dynamics (R(2)=0.37). The overall results show the potential to develop a model-based sensor to monitor ventricular-arterial coupling in clinical real-time.

A multi-scale cardiovascular system model can account for the load-dependence of the end-systolic pressure-volume relationship

Background

The end-systolic pressure-volume relationship is often considered as a load-independent property of the heart and, for this reason, is widely used as an index of ventricular contractility. However, many criticisms have been expressed against this index and the underlying time-varying elastance theory: first, it does not consider the phenomena underlying contraction and second, the end-systolic pressure volume relationship has been experimentally shown to be load-dependent.

Methods

In place of the time-varying elastance theory, a microscopic model of sarcomere contraction is used to infer the pressure generated by the contraction of the left ventricle, considered as a spherical assembling of sarcomere units. The left ventricle model is inserted into a closed-loop model of the cardiovascular system. Finally, parameters of the modified cardiovascular system model are identified to reproduce the hemodynamics of a normal dog.

Results

Experiments that have proven the limitations of the time-varying elastance theory are reproduced with our model: (1) preload reductions, (2) afterload increases, (3) the same experiments with increased ventricular contractility, (4) isovolumic contractions and (5) flow-clamps. All experiments simulated with the model generate different end-systolic pressure-volume relationships, showing that this relationship is actually load-dependent. Furthermore, we show that the results of our simulations are in good agreement with experiments.

Conclusions

We implemented a multi-scale model of the cardiovascular system, in which ventricular contraction is described by a detailed sarcomere model. Using this model, we successfully reproduced a number of experiments that have shown the failing points of the time-varying elastance theory. In particular, the developed multi-scale model of the cardiovascular system can capture the load-dependence of the end-systolic pressure-volume relationship.

Beat-to-beat estimation of the continuous left and right cardiac elastance from metrics commonly available in clinical settings

INTRODUCTION

Functional time-varying cardiac elastances (FTVE) contain a rich amount of information about the specific cardiac state of a patient. However, a FTVE waveform is very invasive to directly measure, and is thus currently not used in clinical practice. This paper presents a method for the estimation of a patient specific FTVE, using only metrics that are currently available in a clinical setting.

METHOD

Correlations are defined between invasively measured FTVE waveforms and the aortic and pulmonary artery pressures from 2 cohorts of porcine subjects, 1 induced with pulmonary embolism, the other with septic shock. These correlations are then used to estimate the FTVE waveform based on the individual aortic and pulmonary artery pressure waveforms, using the "other" dysfunction's correlations as a cross validation.

RESULTS

The cross validation resulted in 1.26% and 2.51% median errors for the left and right FTVE respectively on pulmonary embolism, while the septic shock cohort had 2.54% and 2.90% median errors.

CONCLUSIONS

The presented method accurately and reliably estimated a patient specific FTVE, with no added risk to the patient. The cross validation shows that the method is not dependent on dysfunction and thus has the potential for generalisation beyond pulmonary embolism and septic shock.

Algorithmic processing of pressure waveforms to facilitate estimation of cardiac elastance

Background

Cardiac elastances are highly invasive to measure directly, but are clinically useful due to the amount of information embedded in them. Information about the cardiac elastance, which can be used to estimate it, can be found in the downstream pressure waveforms of the aortic pressure (P_ao) and the pulmonary artery (P_pa). However these pressure waveforms are typically noisy and biased, and require processing in order to locate the specific information required for cardiac elastance estimations. This paper presents the method to algorithmically process the pressure waveforms.

Methods

A shear transform is developed in order to help locate information in the pressure waveforms. This transform turns difficult to locate corners into easy to locate maximum or minimum points as well as providing error correction.

Results

The method located all points on 87 out of 88 waveforms for P_pa, to within the sampling frequency. For P_ao, out of 616 total points, 605 were found within 1%, 5 within 5%, 4 within 10% and 2 within 20%.

Conclusions

The presented method provides a robust, accurate and dysfunction-independent way to locate points on the aortic and pulmonary artery pressure waveforms, allowing the non-invasive estimation of the left and right cardiac elastance.

Mathematical multi-scale model of the cardiovascular system including mitral valve dynamics. Application to ischemic mitral insufficiency

Background

Valve dysfunction is a common cardiovascular pathology. Despite significant clinical research, there is little formal study of how valve dysfunction affects overall circulatory dynamics. Validated models would offer the ability to better understand these dynamics and thus optimize diagnosis, as well as surgical and other interventions.

Methods

A cardiovascular and circulatory system (CVS) model has already been validated in silico, and in several animal model studies. It accounts for valve dynamics using Heaviside functions to simulate a physiologically accurate "open on pressure, close on flow" law. However, it does not consider real-time valve opening dynamics and therefore does not fully capture valve dysfunction, particularly where the dysfunction involves partial closure. This research describes an updated version of this previous closed-loop CVS model that includes the progressive opening of the mitral valve, and is defined over the full cardiac cycle.

Results

Simulations of the cardiovascular system with healthy mitral valve are performed, and, the global hemodynamic behaviour is studied compared with previously validated results. The error between resulting pressure-volume (PV) loops of already validated CVS model and the new CVS model that includes the progressive opening of the mitral valve is assessed and remains within typical measurement error and variability. Simulations of ischemic mitral insufficiency are also performed. Pressure-Volume loops, transmitral flow evolution and mitral valve aperture area evolution follow reported measurements in shape, amplitude and trends.

Conclusions

The resulting cardiovascular system model including mitral valve dynamics provides a foundation for clinical validation and the study of valvular dysfunction in vivo. The overall models and results could readily be generalised to other cardiac valves.

Patient specific identification of the cardiac driver function in a cardiovascular system model

The cardiac muscle activation or driver function, is a major determinant of cardiovascular dynamics, and is often approximated by the ratio of the left ventricle pressure to the left ventricle volume. In an intensive care unit, the left ventricle pressure is usually never measured, and the left ventricle volume is only measured occasionally by echocardiography, so is not available real-time. This paper develops a method for identifying the driver function based on correlates with geometrical features in the aortic pressure waveform. The method is included in an overall cardiovascular modelling approach, and is clinically validated on a porcine model of pulmonary embolism. For validation a comparison is done between the optimized parameters for a baseline model, which uses the direct measurements of the left ventricle pressure and volume, and the optimized parameters from the approximated driver function. The parameters do not significantly change between the two approaches thus showing that the patient specific approach to identifying the driver function is valid, and has potential clinically.

Subject-specific cardiovascular system model-based identification and diagnosis of septic shock with a minimally invasive data set: animal experiments and proof of concept

A cardiovascular system (CVS) model and parameter identification method have previously been validated for identifying different cardiac and circulatory dysfunctions in simulation and using porcine models of pulmonary embolism, hypovolemia with PEEP titrations and induced endotoxic shock. However, these studies required both left and right heart catheters to collect the data required for subject-specific monitoring and diagnosis-a maximally invasive data set in a critical care setting although it does occur in practice. Hence, use of this model-based diagnostic would require significant additional invasive sensors for some subjects, which is unacceptable in some, if not all, cases. The main goal of this study is to prove the concept of using only measurements from one side of the heart (right) in a 'minimal' data set to identify an effective patient-specific model that can capture key clinical trends in endotoxic shock. This research extends existing methods to a reduced and minimal data set requiring only a single catheter and reducing the risk of infection and other complications-a very common, typical situation in critical care patients, particularly after cardiac surgery. The extended methods and assumptions that found it are developed and presented in a case study for the patient-specific parameter identification of pig-specific parameters in an animal model of induced endotoxic shock. This case study is used to define the impact of this minimal data set on the quality and accuracy of the model application for monitoring, detecting and diagnosing septic shock. Six anesthetized healthy pigs weighing 20-30 kg received a 0.5 mg kg(-1) endotoxin infusion over a period of 30 min from T0 to T30. For this research, only right heart measurements were obtained. Errors for the identified model are within 8% when the model is identified from data, re-simulated and then compared to the experimentally measured data, including measurements not used in the identification process for validation. Importantly, all identified parameter trends match physiologically and clinically and experimentally expected changes, indicating that no diagnostic power is lost. This work represents a further with human subjects validation for this model-based approach to cardiovascular diagnosis and therapy guidance in monitoring endotoxic disease states. The results and methods obtained can be readily extended from this case study to the other animal model results presented previously. Overall, these results provide further support for prospective, proof of concept clinical testing with humans.

Model-based prediction of the patient-specific response to adrenaline

A model for the cardiovascular and circulatory systems has previously been validated in simulated cardiac and circulatory disease states. It has also been shown to accurately capture the main hemodynamic trends in porcine models of pulmonary embolism and PEEP (positive end-expiratory pressure) titrations at different volemic levels. In this research, the existing model and parameter identification process are used to study the effect of different adrenaline doses in healthy and critically ill patient populations, and to develop a means of predicting the hemodynamic response to adrenaline. The hemodynamic effects on arterial blood pressures and stroke volume (cardiac index) are simulated in the model and adrenaline-specific parameters are identified. The dose dependent changes in these parameters are then related to adrenaline dose using data from studies published in the literature. These relationships are then used to predict the future, patient-specific response to a change in dose or over time periods from 1-12 hours. The results are compared to data from 3 published adrenaline dosing studies comprising a total of 37 data sets. Absolute percentage errors for the identified model are within 10% when re-simulated and compared to clinical data for all cases. All identified parameter trends match clinically expected changes. Absolute percentage errors for the predicted hemodynamic responses (N=15) are also within 10% when re-simulated and compared to clinical data. Clinically accurate prediction of the effect of inotropic circulatory support drugs, such as adrenaline, offers significant potential for this type of model-based application. Overall, this work represents a further clinical, proof of concept, of the underlying fundamental mathematical model, methods and approach, as well as providing a template for using the model in clinical titration of adrenaline in a decision support role in critical care. They are thus a further justification in support of upcoming human clinical trials to validate this model.

Unique parameter identification for cardiac diagnosis in critical care using minimal data sets

Lumped parameter approaches for modelling the cardiovascular system typically have many parameters of which a significant percentage are often not identifiable from limited data sets. Hence, significant parts of the model are required to be simulated with little overall effect on the accuracy of data fitting, as well as dramatically increasing the complexity of parameter identification. This separates sub-structures of more complex cardiovascular system models to create uniquely identifiable simplified models that are one to one with the measurements. In addition, a new concept of parameter identification is presented where the changes in the parameters are treated as an actuation force into a feed back control system, and the reference output is taken to be steady state values of measured volume and pressure. The major advantage of the method is that when it converges, it must be at the global minimum so that the solution that best fits the data is always found. By utilizing continuous information from the arterial/pulmonary pressure waveforms and the end-diastolic time, it is shown that potentially, the ventricle volume is not required in the data set, which was a requirement in earlier published work. The simplified models can also act as a bridge to identifying more sophisticated cardiac models, by providing an initial set of patient specific parameters that can reveal trends and interactions in the data over time. The goal is to apply the simplified models to retrospective data on groups of patients to help characterize population trends or un-modelled dynamics within known bounds. These trends can assist in improved prediction of patient responses to cardiac disturbance and therapy intervention with potentially smaller and less invasive data sets. In this way a more complex model that takes into account individual patient variation can be developed, and applied to the improvement of cardiovascular management in critical care.

Current progress in patient-specific modeling

We present a survey of recent advancements in the emerging field of patient-specific modeling (PSM). Researchers in this field are currently simulating a wide variety of tissue and organ dynamics to address challenges in various clinical domains. The majority of this research employs three-dimensional, image-based modeling techniques. Recent PSM publications mostly represent feasibility or preliminary validation studies on modeling technologies, and these systems will require further clinical validation and usability testing before they can become a standard of care. We anticipate that with further testing and research, PSM-derived technologies will eventually become valuable, versatile clinical tools.