Date of Award

5-2020

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Mechanical Engineering

Committee Chair/Advisor

Dr. Ethan Kung, Committee Chair

Committee Member

Dr. Georges Fadel

Committee Member

Dr. William Richardson

Abstract

Lumped Parameter Networks (LPN) are computational models that can capture the systemic responses of the cardiovascular system. An LPN describes the systemic behavior by providing outputs such as the blood pressure, flowrate, and heart chamber volume at the points defined in the network. To use the LPN for medical applications, it needs to provide output values that represent the physiology of realistic subjects. The cardiovascular system exhibits short-term and long-term adaptations to factors such as aging, weight gain, exercise, etc. Therefore, subject characteristics such as age, height, weight, etc. affect the output values of the LPN. Current literature reports the mean reference values of the outputs for the population and the trends of the outputs for each characteristic separately. However, the combined effects of multiple characteristics and the corresponding numerical output values are generally not reported.

This study first remodels an existing LPN to represent an average healthy adult. Next, a systematic review was conducted to compile blood pressure, cardiac output, and ventricular volume data for groups of healthy subjects. Regression algorithms used this aggregate data to formulate predictive models for the outputs – systolic and diastolic blood pressure, ventricular volumes, cardiac output, and heart rate – against the characteristics – age, height, weight, and exercise intensity. A simulation-based procedure generated data of virtual subjects to confirm whether these regression models built based on aggregate data can perform well for individual-level predictions. The blood pressure and heart rate models were also validated using real individual data.

Although the multivariable regression models use aggregate data, the expected error for individual predictions is provided. The direction of trends between model outputs and the input subject characteristics reported in this study agree with the results in current literature. The high prediction errors (20% - 50%) can be attributed to the best models being linear. Although other studies observe exponential predictor-output relations, the non-linear algorithms do not fit the data well.

Future studies involving the LPN can use the output values predicted by these regression models as targets to model the cardiovascular system of a population cohort or an individual described using age, height, weight, and exercise intensity.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.