Real time segmentation of heart sounds
| dc.contributor.advisor | Booysen, M. J. | en_ZA |
| dc.contributor.author | Fourie, David | en_ZA |
| dc.contributor.other | Stellenbosch University. Faculty of Engineering. Dept. of Electrical and Electronic Engineering. | en_ZA |
| dc.date.accessioned | 2015-12-14T07:43:00Z | |
| dc.date.available | 2015-12-14T07:43:00Z | |
| dc.date.issued | 2015-12 | |
| dc.description | Thesis (MEng)--Stellenbosch University, 2015. | en_ZA |
| dc.description.abstract | ENGLISH ABSTRACT: The poor state of the healthcare system in South Africa has resulted in unacceptable high levels of infant mortality. Congenital heart disease is one of the main contributions to these high rates of mortality, with the cost of treatment and the availability of specialists being the driving factors. Computer aided auscultation is a technological solution to assist with the diagnosis of the disease. In its current form, computer aided auscultation is unsuitable for continuous patient monitoring. The aim of this thesis is to develop an algorithm that will allow the existing methods of computer aided auscultation to work in real time so they can be used in patient monitoring. Existing methods of identifying the first and second heart sound are limited to offline processing. The algorithm developed in this thesis uses the correlation of the time-frequency coefficients of individual heart sounds to generate a feature vector for each heart sound that can be used to separate the sounds into different groups. To test the performance of the algorithm, 230 heart sounds from normal patients were first manually segmented and then processed with the algorithm. The noise sensitivity of the algorithm was also tested using generated heart sounds. Finally, the real time capability of the algorithm was tested. The testing against sounds for normal patients resulted in a 84.2 % accuracy and an 84.4% hit rate. The synthetic testing showed the system starts to perform badly with a signal to noise ratio lower than -10db. The real time testing of the system showed that the algorithm is fast enough to be used in a real time environment. This thesis concludes that proposed algorithm is suitable for the detection of the first and second heart sounds in real time. | en_ZA |
| dc.description.abstract | AFRIKAANSE OPSOMMING: Die toestand van die gesondheidstelsel in Suid-Afrika lei tot onaanvaarbare vlakke van kindersterftes. Oorerflike hartksiektes is een van die hoofoorsake van hierdie sterftesyfers, aangedryf deur die koste van behandeling en die tekort aan beskikbaarheid van spesialiste. Rekenaargesteundebeluistering is ’n tegnologiese oplossing wat help met die diagnose van hierdie kwaal. Huidiglik is rekenaargesteundebeluistering ongeskik vir aaneenlopende pasientemonitering. Die doelwit van hierdie tesis in om ’n algoritme te onwikkel wat sal toelaat dat bestaande metodes van rekenaargestuendebeluistering intyds sal werk sodat gebruik kan word vir deurlopende monitering van pasiente. Bestaande metodes, wat aangewend word om die eerste- en tweede hartklanke te identifiseer, is beperk tot nie-intydse verwerking. Die algoritme wat in hierdie tesis ontwikkel is, gebruik die korrelasie van die tyd-frekwensie koeffisiente van individuele hartklanke om ’n eienskapsvektor vir elke hartklank te genereer, wat dan gebruik word om die hartklanke in verskillende groepe in te deel. Om die werkverrigting van die algoritme te toets, is 230 hartklanke van pasiente met normale harte eers per hand gesegmenteer en daarna met die algoritme verwerk. Die algoritme se bestandheid teen ruis is ook getoets deur gebruik te maak van sintetiese hartklanke. Uiteindelik is die intydse vermoë van die algortime getoets. Die toetsing met normale hartklanke het ’n akkuraatheid van 84.2% en trefkoers van 84.4% opgelewer. Die ruisbestandheidstoetse het aangedui dat die stelsel sleg begin werkverrigting verloor met sein-tot-ruis-verhoudings van laer as -10dB. Die intydse toetsing het aangedui dat die algoritme vining genoeg is om gebruik te word in ’n intydse implementering. Die tesis maak die gevolgtrekking dat die voorgestelde algoritme geskik is vir die intydse identifisering van die eerste en tweede hartklanke. | af_ZA |
| dc.format.extent | 69 pages : illustrations | en_ZA |
| dc.identifier.uri | http://hdl.handle.net/10019.1/97891 | |
| dc.language.iso | en_ZA | en_ZA |
| dc.publisher | Stellenbosch : Stellenbosch University | en_ZA |
| dc.rights.holder | Stellenbosch University | en_ZA |
| dc.subject | Computer aided auscultation | en_ZA |
| dc.subject | Heart sounds -- Segmentation | en_ZA |
| dc.subject | UCTD | en_ZA |
| dc.title | Real time segmentation of heart sounds | en_ZA |
| dc.type | Thesis | en_ZA |