doktorska disertacija
Vojko Glaser (Author), Aleš Holobar (Mentor)

Abstract

Analiza elektromiogramov (EMG) je v medicini izredno pomembna in pogosto predstavlja ključen del diagnostike. V Laboratoriju za sistemsko programsko opremo UM FERI je bila v preteklosti razvita metoda kompenzacije konvolucijskih jeder (ang. convolution kernel compensation - CKC), ki signale EMG dekomponira na prispevke posameznih motoričnih enot. Ta metoda je bila obširno testirana in se je v primeru izometričnih mišičnih skrčitev izkazala za izredno učinkovito. V primeru dinamičnih mišičnih skrčitev pa nastopijo geometrijske deformacije mišičnega tkiva, ki močno zmanjšajo učinkovitost metode CKC. V doktorski disertaciji podrobno preučimo spremembe površinskih signalov EMG pri dinamičnih mišičnih skrčitvah in predstavimo dva nova algoritma za njihovo dekompozicijo na prispevke motoričnih enot. Oba algoritma izhajata iz metode CKC. Prvi algoritem izkorišča v doktorski disertaciji predstavljeno ugotovitev, da so nepotujoče komponente akcijskih potencialov motoričnih enot (APME) bistveno manj občutljive na geometrijske spremembe mišice kot potujoče komponente APME in metodo CKC prilagodi zaznavi nepotujočih komponent APME. Drugi postopek temelji na dinamični obtežitvi prispevkov posameznih motoričnih enot v izmerjenih signalih EMG in s tem omogoči avtomatsko prilagajanje metode CKC dinamičnim spremembam APME. Oba algoritma smo ovrednotili s sintetičnimi in z eksperimentalnimi površinskimi signali EMG. V ta namen smo obstoječi napredni simulator površnikih signalov EMG funkcionalno dogradili tako, da simulira dinamične mišične skrčitve in z njimi analizira vpliv šuma, prostorskih filtrov in razpona dinamičnih skrčitev na učinkovitost obeh predstavljenih algoritmov. Eksperimentalne signale smo izmerili nad mišicama vastus lateralis in rectus femoris pri petih zdravih mladih preiskovancih. Izmerili smo dve hitrosti upogiba kolenskega sklepa, in sicer 5 °/s in 10 °/s. V vseh testih se je za najbolj učinkovito izkazala metoda z dinamično obtežitvijo prispevkov motoričnih enot, ki je v primeru mišice vastus lateralis z visoko natančnostjo razpoznala 6,5 % 1,8, v primeru mišice rectus femoris pa 4,5 % 1,6 motoričnih enot. Metoda, ki temelji na uporabi nepotujoče komponente APME, je bila statistično značilno manj učinkovita, še zlasti v primeru eksperimentalnih signalov EMG.

Keywords

slepa ločitev signalov;večkanalni površinski elektromiogram;dinamične mišične skrčitve;doktorske disertacije;

Data

Language: Slovenian
Year of publishing:
Typology: 2.08 - Doctoral Dissertation
Organization: UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher: [V. Glaser]
UDC: 004.383.3:57.089:612.741(043.3)
COBISS: 19687958 Link will open in a new window
Views: 947
Downloads: 130
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Blind Separation of Multichannel Surface Electromyograms During Dynamic Muscle Contractions
Secondary abstract: Analysis of electromyograms (EMG) is of great importance in medicine as it often represents a key part of medical diagnostics. In the last decade, the method called Convolution Kernel Compensation (CKC) was developed in System Software Laboratory at Faculty of Electrical Engineering and Computer Science, University of Maribor. This method successfully decomposes EMG signals into contributions of individual motor units. It was extensively tested in different experimental conditions and has proven to be very efficient in the case of isometric muscle contractions. However, this is no longer the case in dynamic muscle contractions. The latter introduce geometrical deformations of muscle tissues which drastically affect the CKC method. In this thesis, we study changes of surface EMG in the case of dynamic muscle contractions and introduce two new surface EMG decomposition approaches. Both approaches are based on the CKC method. The first algorithm builds on the notion that measured motor unit action potentials (MUAPs) combine travelling and non-travelling components. We demonstrate that non-travelling component is significantly less prone to geometric deformations of a muscle than its traveling counterpart. For this reason we focus our decomposition on non-travelling component of a MUAP only. The second algorithm is based on dynamical weighting of contributions from individual motor units in CKC and adapts very well to the dynamic changes of MUAPs. Both approaches have been tested on synthetic and experimental surface EMG signals. Synthetic signals have been generated with adapted surface EMG simulator that discretises the muscle contractions and supports systematic testing of noise impact, effect of spatial filters and different speeds and ranges of dynamic contractions. Experimental surface EMG signals have been obtained from vastus lateralis and rectus femoris muscles of five young healthy subjects during knee bending at speeds of 5°/s and 10°/s, respectively. The best results were obtained with dynamical weighting of motor unit contributions. This method detected with high reliability 6.5 % 1.8 and 4.5 % 1.6 motor units in vastus lateralis and rectus femoris muscle, respectively. The method using non-travelling MUAP component was significantly less efficient, especially in the case of experimental EMG signals.
Secondary keywords: blind separation;multichannel surface electromiograms;dynamic muscle contractions;
URN: URN:SI:UM:
Type (COBISS): Dissertation
Thesis comment: Univ. v Mariboru, Fak. za elektrotehniko, računalništvo in informatiko
Pages: [VI], 118 str.
ID: 9143180