1. UKM Learning and Research Repository
2. PHD Thesis / Tesis PHD
3. Faculty of Engineering and Built Environment / Fakulti Kejuruteraan dan Alam Bina

Title:	Photoplethysmography flow mediated dilation based endothelial dysfunction assessment using fuzzy logic and neural network
Authors:	Mojgan Zaheditochai (P342484)
Supervisor:	Mohd Alauddin Mohd Ali, Prof. Dr.
Keywords:	Neural networks (Computer science) Vascular endothelial cells Plethysmography Blood flow -- Measurement Biomedical engineering Fuzzy logic Universiti Kebangsaan Malaysia -- Dissertations Dissertations, Academic -- Malaysia
Issue Date:	14-Aug-2011
Description:	Endothelial dysfunction (ED) is considered as a major cause of the development of atherosclerosis (hardening and thickening of the arterial blood vessel walls) and ultrasound flow mediated vasodilation (US-FMD) is used to derive a non-invasive index to measure ED. In this technique, high frequency ultrasonography (US FMD) images of the brachial artery are recorded to estimate brachial artery diameter changes. Although this technique has yielded important information about vascular function, it requires a skilled operator, five minutes blood flow blockage that is uncomfortable to the subject and is limited to vessel sizes of more than 2.5 millimeter. Alternatively the method selected for this work is based on peripheral pulse analysis through photoplethysmography (PPG FMD) which is low-cost, non-invasive and simple to use. The main goal of this thesis is to identify new features of PPG FMD and US FMD responses to enable the use of PPG FMD instead of US FMD for vascular disease diagnosis. Firstly, PPG and US FMD signals are preprocessed. For the PPG FMD, DC removal, peak detection, low pass filtering, baseline identification, normalization based on the baseline, extracting curve versus time are implemented. Similarly for the US FMD, interpolation, curve fitting, baseline definition, normalization based on the baseline, extracting the curve versus time are implemented. Secondly, new features based on the time and area under the curve are extracted from PPG and US FMD. Then fuzzy analysis is applied on features as well as heart risk factors. Finally, the fuzzified PPG FMD features and risk factors (correction factors) are applied to a feed forward back propagation neural network where the fuzzified US FMD features are used as target outputs. The regression (r-value) between real output of the network (referring to the PPG FMD) and target output (referring to the US FMD) is an index to show the correlation of the two methods. The database which was collected in a previous study consists of records from 125 subjects. These are used for the above neural network, where 70% are for training data and 30% for test data. The postregression processing of the output and target of the neural network showed the correlation between PPG FMD and US FMD responses is 77% for test data. The extracted features show significant separation between healthy (without a specific risk factor) and pathological group (at least having either risk factor of obesity (N=79), hyperlipidemia (N=16) or diabetes (N=11)) by p-value of 0.0001. Therefore, the more elaborate features when combined with the data processing techniques enable the PPG FMD, with the addition of correction factors, to be used instead of US FMD for assessing ED.,"Certification of Master's/Doctoral Thesis" is not available,Ph.D.
Pages:	116
Call Number:	TA164.Z338 2012 3 tesis
Publisher:	UKM, Bangi
Appears in Collections:	Faculty of Engineering and Built Environment / Fakulti Kejuruteraan dan Alam Bina

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.