In-silico cardiovascular hemodynamic model to simulate the effect of physical exercise

Highlights

• We propose a model to derive cardiac insights during physical activities of a person.

• To run the model, heart variables are assessed from ECG data to replicate the activation delays.

• This approach is more realistic than traditional techniques.

• Effects of exercise are simulated using two datasets to assess cardiac parameters.

• Model-simulated BPs are highly correlated with the ground truth BPs of Kaggle data.

Abstract

Coronary heart disease (CHD) is a leading cause of death in developing countries. Lifestyle disorders are one of the precursors of CHD. Exercise-based cardiac health rehabilitation is often prescribed as a secondary prevention strategy to reduce the impact of CHD. To monitor the exercise’s impact on the cardiac parameters, a simulation platform is required to predict the cardiovascular parameters of an individual during exercise. So, in this paper, our objective is to model an in-silico cardiac platform to simulate the exercise effect on the cardiac parameters relevant to cardiac rehabilitation. The proposed framework encompasses a novel cardiovascular hemodynamic model for simulating an individual’s exercise condition using an ECG signal and the metabolic equivalent of task (MET) value approximated from a wearable device. The Neuromodulation parameters such as heart rate and cardiac compliances are estimated from the morphology of the ECG signal. The exercise modulated systemic resistances for the coupled hemodynamic model are assessed based on the MET values of the activity levels to meet additional blood demand during exercises. Effects of exercise on cardiac parameters are simulated on two open-source datasets for healthy subjects, namely Troika and Kaggle, and several hemodynamic parameters are evaluated. The model simulated blood pressure (BP) shows a correlation coefficient of 0.99 (for systolic BP) and 0.95 (for diastolic BP) with Kaggle data BP. The effect of exercise on a diseased heart is demonstrated through a case of silent ischemia. The proposed simulation platform is intended to be used for cardiac rehabilitation monitoring aid for both healthy and patients suffering from heart diseases, providing a cardiac care continuum.

Introduction

Cardiovascular diseases (CVD) are the main cause of death worldwide with coronary heart disease (CHD) accounting for the majority of CVD mortality [1]. CHD has a high prevalence and is aggravated by lifestyle disorders [2], [3]. To reduce the impact of CHD [4], [5], exercise-based cardiac rehabilitation (CR) is often prescribed, which aims to optimize cardiovascular disease risk reduction, promote the adoption and adherence of healthy habits [6]. Thus, CR is prescribed to patients suffering from cardiac diseases like valvular heart disease, heart transplantation, heart failure with reduced ejection fraction (EF), post-coronary artery bypass grafting (CABG), etc. to improve the quality of life and reduce re-hospitalization [2]. Furthermore, according to the retrospective reports [7], a regular physical workout is substantially linked to a lower risk of cardiovascular (CV) mortality and morbidity, and it can regulate several established CHD risk factors like blood pressure, blood lipid profile, glucose metabolism, weight status, and body composition through cardiovascular and metabolic adaptation [8]. Additionally, exercise is standardized as a part of illness prevention and treatment, and the exercise prescriptions are widely acceptable in other clinical medicine domains, the most recent being COVID-19 treatment [9]. Intelligent regulation of human exercise behaviors has become significantly necessary for exercise medicine after the COVID-19 epidemic. Exercise prescription is based on the exercise intensity that is often detected by Oxygen Uptake ($V{O}_2$) and Heart Rate (HR) [1]. Along with exercise medicine, the response of the heart under stress should be accounted for and evaluated in CDH patient management plans. Exercise regimens are mostly pseudo-close-loop where exercise physiological responses are not reliably constrained by the uncertainty in pathological conditions. Most of the stress and exercise conditions create a high demand on the cardiac system, often revealing adverse conditions during exercise. Recent statistics from WHO [9] report that around 80000 athletes died in sudden cardiac arrest in the last year. Hence, for CR or exercise prescriptions to be fruitful, modeling and monitoring exercise behavior on cardiac parameters are utterly required.

To address these issues, a cardiac simulation framework is required for predicting cardiopulmonary parameters and infusing predictability and intelligence in the exercise medicine prescription of an individual during exercise. The developed tool should be capable of measuring the efficiency of an individual during exercise or daily living activities. Additionally, such a system could also be useful for recording cardiac irregularities and generating alarms, aiding in the early screening of several CHD.

Computer simulation-based modeling in exercise health care is an attractive proposition. Mathematical models are often used to improve understanding of exercise physiology which in turn is useful for predicting adverse accidents like sudden cardiac death during exercise [10]. Clinicians could use such predictive models to stratify the likelihood or severity of exercise intolerance in their patients. Such platforms could also act as virtual test beds to verify the consequences of different levels of exercise for pathological conditions of varying severity.

There has been various literature defining the cardiac parameter variations based on lumped order models [11]. Contemporary computational fluid dynamics methods offer a powerful approach to analyzing patient-specific hemodynamic conditions. There exist some simulation platforms [11], [12], [13], [14] to generate the effect of exercises; however, all of them are dependent on the Neuromodulation scheme (baroreflex autoregulation [13]). So, these models are based on the assumption that the threshold parameters of Neuromodulation correlate with physical activities. So, to run these models, subject-specific prior knowledge of the threshold parameters would be required. As a consequence, such models may not work practically for a large set of populations. Moreover, the traditional physiological signals (such as electrocardiogram (ECG) and/or photoplethysmography (PPG))-based blood-pressure (BP) measurement techniques [15], [16], [17], [18], [19] can only estimate the systolic and diastolic BPs, no other cardiac parameters such as cardiac output, ejection fraction, etc. could be evaluated.

Apart from exercise modeling, exercise monitoring programs are also being actively researched and implemented [20]. The adaptation of smart wearable devices has resulted in a significant increase in cardiac health monitoring [20]. Smart wearable gadgets, unlike traditional medical devices, allow for 24/7 monitoring and recording of physiological data. Wearable monitoring mostly aims at providing simple cardiac indices that are proven to be useful biomarkers for early screening of CVD, like heart rate (HR), activity recognition, and calorie consumption [21]. HR is frequently measured using electrocardiogram (ECG) and photoplethysmogram (PPG), with the latter being the most prevalent method with wearable devices [22].

Taking a clue from the requirement of both exercise monitoring and modeling for establishing a cardiac care continuum for cardiac rehabilitation, we propose an in-silico cardiovascular digital-twin simulation framework (Fig. 1) to derive hemodynamic insights of exercise and ambulatory activities of a person employing the real-time ECG signal (while performing physical activities). The cardiovascular model is demonstrated by the pressure–volume dynamics of cardiac chambers, as well as the systemic and pulmonic circulations. Several cardiac parameters such as heart rate and cardiac compliance are extracted from the recorded ECG data to simulate the model. The compliance parameters are derived from the morphology of the ECG signal; hence, the activation delays in the heart chambers are reproduced purposefully [23]. This specific approach is more realistic than the threshold based Neuromodulation techniques [11], [12], [13], [14], so, this model can be employed on a wide range of people. To model the systemic resistance for a task-specific activity, the metabolic equivalent of the task or the MET level has been employed. For cases where MET levels are not defined like ambulatory activities, an algorithm has been developed to determine the MET level from wearable accelerometer data. A novel addition is the inclusion of unstressed blood volume in circulation, which plays an important role in exercise physiology [24]. The rate of change of unstressed blood volume has been coupled with the pulmonary arterial pressure, so, during intensive physical activities, additional blood demand could be met by increasing the venous return during the diastolic phase. The model is also capable of generating continuous blood pressure data during exercise simulation by only employing ECG data. Apart from healthy exercise condition simulation, the effect of exercise under pathological conditions has also been demonstrated using the test case of Silent ischemia [25]. From the simulated hemodynamic parameters, the effect of exercise intensity on changes in cardiac parameters like ejection fraction are evident, making the proposed model suitable for employ in exercise simulation for both pathological and non-pathological conditions. It is to be noted that in our initial study [26], we have provided a hemodynamic framework to simulate the effect of exercise using ECG data. In that study, there is no option for personalization, and additional blood volume response during the ventricular diastolic phase has not been considered. In this paper, we have extended that idea by developing a personalized model that includes the unstressed blood volume response, systemic resistance estimation from MET values, and other features. Moreover, the manifestation of silent pathophysiological conditions aggravated by exercise has also been reported here. One important contribution of the current work is that we have extensively validated the simulated BPs to the ground-truth BPs of the Kaggle data. Our comparison study shows the efficacies of the proposed approach.

The rest of the paper is organized as follows: In Section 2, we have introduced the basics of the cardiovascular model along with the estimation of the compliance parameters. The modeling of the systemic resistance to the MET level during exercise has been discussed in Section 3. Section 4 represents the simulation results of the proposed system with two well-known datasets. The validation, comparison with other techniques, and the diseased heart simulation are displayed in Section 5. Finally, Section 6 concludes the work.