Abstract

Systolic time intervals (STI) are clinically used as non-invasive predictor of cardiovascular disease. However, algorithm accuracy generally suffers across subjects and physiological states, requiring parameter tuning for robust STI extraction. To address this challenge, an automated methodology of processing with varying tuning parameters was explored. In this work, two STIs were examined: the R-wave pulse transit time to the PPG foot at the ear (rPTT) and the left ventricular ejection time (LVET).

Historic feature detection algorithms were used with a range of tuning parameters over a 60 second interval, with least variance used to select the optimal parameter for robust extraction. These least variance algorithms were quantitatively compared to historic, single parameter algorithms using a positive predictive value metric. In order to decrease the runtime of the algorithms, the least variance algorithms were written such that they could run on a GPU using CUDA.

Overall, the least variance algorithms were able to extract the features better than the historic algorithms, without sacrificing runtime. In addition to providing this robust and reliable STI extraction, the least variance algorithms can be adapted to extract features from any period data stream.

Library of Congress Subject Headings

Heart beat--Mathematical models; Algorithms

Publication Date

5-21-2014

Document Type

Thesis

Student Type

Graduate

Degree Name

Computer Engineering (MS)

Department, Program, or Center

Microsystems Engineering (KGCOE)

Advisor

David A. Borkholder

Advisor/Committee Member

Roy Melton

Advisor/Committee Member

Andres Kwasinski

Comments

Physical copy available from RIT's Wallace Library at RC681 .C94 2014

Campus

RIT – Main Campus

Plan Codes

CMPE-MS

Download

Share

COinS