(A) knowledge-based approach to pulse feature extraction and design of methodology for arterial stiffness diagnosis

Abstract

Previous studies have suggested that subjects with cardiovascular disease (CVD), which is one of the highest mortality rates in Korea, have increased arterial stiffness compared with subjects without CVD. From this relationship, many researchers have investigated arterial stiffness to indirectly estimate cardiovascular events. In recent years, various measurement techniques have been proposed and applied for assessing arterial stiffness: PWV, Ultrasonography, MRI, etc. Although these techniques show high performance in assessment, they are cost expensive and have poor usability. This thesis, therefore, has applied pulse contour analysis (PCA) technique to screen out these disadvantages. Furthermore, this thesis has suggested a knowledge-based approach to accurate pulse feature extraction from both PTG and SDPTG. Although most of the previous researches made various contributions to the automatic extraction of pulse parameters, their achievements have limitations in applying the algorithms to various pulse shapes. In order to solve this problem, this thesis has extracted pulse features by using the knowledge derived from several previous pulse periods. This approach can improve feature extraction rates and reduce false positive detections generated by motion artifacts. In order to examine the relationship between baPWV, which is widely used for arterial stiffness evaluation, and various pulse features, this thesis has performed linear regression analysis by using a SPSS tool. Before analysis, it is required that time-related pulse features, such as reflected wave arrival time, must be corrected by heart rate (or peak to peak time interval). Since these parameters are directly affected by heart rate, this thesis has applied rate correction algorithms which are commonly used in ECG analysis - linear regression equation, Bazett``s formula, and Fridericia``s formula - to the time-related pulse parameters. From the results of statistical comparison, time-related ...