Purpose The goal is to develop a cardiovascular virtual patient database (VPD) combining physiological and demographic data to provide the foundation for future applications in medical diagnostics, decision-making, credibility testing, and formal uncertainty analysis, and to enable its integration with three-dimensional (3D) hemodynamic models and to train neural networks. Methods We generate an initial VPD by treating input parameters of a low-dimensional cardiovascular model as stochastic variables. Literature data and sensitivity analysis ensured physiological plausibility, while resampling improved physiological accuracy. Key physiological quantities are included such as systolic and diastolic aortic pressure, radial and carotid pressure, cardiac output, and diagnostic pulse wave velocities. Demographic factors (sex and age) are assigned based on their physiological impact. The open-source hemodynamic solver, first_blood, ensures accuracy and low computational time. Results The initial VPD consists of 50,000 Virtual Patients; after resampling, 34,347 remain in the final VPD. The difference of diastolic and systolic aortic pressures between the VPD (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">$70.24\pm 14.3$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>70.24</mn> <mo>±</mo> <mn>14.3</mn> </mrow> </math></EquationSource> </InlineEquation> and <InlineEquation ID="IEq2"> <EquationSource Format="TEX">$116.8\pm 16.11$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>116.8</mn> <mo>±</mo> <mn>16.11</mn> </mrow> </math></EquationSource> </InlineEquation> mmHg) and the literature (<InlineEquation ID="IEq3"> <EquationSource Format="TEX">$75.6\pm 12.7$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>75.6</mn> <mo>±</mo> <mn>12.7</mn> </mrow> </math></EquationSource> </InlineEquation> and <InlineEquation ID="IEq4"> <EquationSource Format="TEX">$113.0\pm 11.2$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>113.0</mn> <mo>±</mo> <mn>11.2</mn> </mrow> </math></EquationSource> </InlineEquation> mmHg) is low. The differences caused by the sex of the patient are reproduced well by the VPD: increased diastolic aortic pressure for males (<InlineEquation ID="IEq5"> <EquationSource Format="TEX">$72.1\pm 12.6$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>72.1</mn> <mo>±</mo> <mn>12.6</mn> </mrow> </math></EquationSource> </InlineEquation> and <InlineEquation ID="IEq6"> <EquationSource Format="TEX">$68.5\pm 15.4$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>68.5</mn> <mo>±</mo> <mn>15.4</mn> </mrow> </math></EquationSource> </InlineEquation> for males and females respectively). The VPD also accurately includes higher pulse wave velocities with age, patients below year 30 have <InlineEquation ID="IEq7"> <EquationSource Format="TEX">$6.3\pm 0.5$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>6.3</mn> <mo>±</mo> <mn>0.5</mn> </mrow> </math></EquationSource> </InlineEquation> and above 70 have <InlineEquation ID="IEq8"> <EquationSource Format="TEX">$8.8\pm 1.2$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>8.8</mn> <mo>±</mo> <mn>1.2</mn> </mrow> </math></EquationSource> </InlineEquation> m/s with a linear increment in-between. Conclusions The proposed methodology and first_blood solver effectively generate physiologically realistic virtual patient waveforms and demographic variability, providing a robust database for 3D cardiovascular simulations, machine-learning training datasets, and potential clinical decision support applications

Synthetic, Population-Based Virtual Patient Database Using a Digital Twin of the Cardiovascular System

期刊论文
出版模式：

开放获取
发表日期：

2026年2月12日

查看全文

Richárd Wéber,
Márta Viharos,
Benjamin Csippa,
Dániel Gyürki,
György Paál

摘要

Purpose

The goal is to develop a cardiovascular virtual patient database (VPD) combining physiological and demographic data to provide the foundation for future applications in medical diagnostics, decision-making, credibility testing, and formal uncertainty analysis, and to enable its integration with three-dimensional (3D) hemodynamic models and to train neural networks.

Methods

We generate an initial VPD by treating input parameters of a low-dimensional cardiovascular model as stochastic variables. Literature data and sensitivity analysis ensured physiological plausibility, while resampling improved physiological accuracy. Key physiological quantities are included such as systolic and diastolic aortic pressure, radial and carotid pressure, cardiac output, and diagnostic pulse wave velocities. Demographic factors (sex and age) are assigned based on their physiological impact. The open-source hemodynamic solver, first_blood, ensures accuracy and low computational time.

Results

The initial VPD consists of 50,000 Virtual Patients; after resampling, 34,347 remain in the final VPD. The difference of diastolic and systolic aortic pressures between the VPD ( $70.24\pm 14.3$ $70.24 \pm 14.3$ and $116.8\pm 16.11$ $116.8 \pm 16.11$ mmHg) and the literature ( $75.6\pm 12.7$ $75.6 \pm 12.7$ and $113.0\pm 11.2$ $113.0 \pm 11.2$ mmHg) is low. The differences caused by the sex of the patient are reproduced well by the VPD: increased diastolic aortic pressure for males ( $72.1\pm 12.6$ $72.1 \pm 12.6$ and $68.5\pm 15.4$ $68.5 \pm 15.4$ for males and females respectively). The VPD also accurately includes higher pulse wave velocities with age, patients below year 30 have $6.3\pm 0.5$ $6.3 \pm 0.5$ and above 70 have $8.8\pm 1.2$ $8.8 \pm 1.2$ m/s with a linear increment in-between.

Conclusions

The proposed methodology and first_blood solver effectively generate physiologically realistic virtual patient waveforms and demographic variability, providing a robust database for 3D cardiovascular simulations, machine-learning training datasets, and potential clinical decision support applications