e-space
Manchester Metropolitan University's Research Repository

The utility of b-mode ultrasound for the diagnosis of motor neurone disease: automated detection and analysis of muscle twitches in ultrasound Images of motor neurone disease affected participants and healthy controls

Bibbings, Kate (2017) The utility of b-mode ultrasound for the diagnosis of motor neurone disease: automated detection and analysis of muscle twitches in ultrasound Images of motor neurone disease affected participants and healthy controls. Doctoral thesis (PhD), Manchester Metropolitan University.

[img]
Preview

Available under License Creative Commons Attribution Non-commercial No Derivatives.

Download (8MB) | Preview

Abstract

Motor Neurone Disease (MND) is a progressive, neurodegenerative disease, for which there is no known cure. Electromyography (EMG) is the standard technique for the detection of diagnostic indicators, such as fasciculations (twitches). Ultrasound (US) imaging may provide a more sensitive alternative to EMG for detection of fasciculations. However, only one computational technique has previously been applied to image sequences to provide an objective measure of fasciculation occurrence. The work presented here therefore describes the development and evaluation of a new computational approach, based on foreground detection using a mixture of Gaussians (GMM). In addition, the only other computational analysis approach available, which is based on feature tracking and mutual information analysis (KLT/MI) was further evaluated. Two data sets were used to evaluate the computational approaches. The first data set had previously been collected and comprised US images from medial gastrocnemius (MG) and biceps brachii (BB) from healthy (n = 20) and MND affected (n = 5) participants. The second data set comprised simultaneously recorded US images and intramuscular EMG from five muscles (medial gastrocnemius (MG), biceps brachii (BB), rectus femoris (RF), trapezius (TRAP), rectus abdominis (RA) and thoracic paraspinal (TP)) of healthy (n = 20) and MND affected (n = 20) participants. Accuracy of the approaches for fasciculation detection was evaluated against two measures of ground-truth: i) manual identification; ii) intramuscular EMG. Accuracy was defined as the area under the receiver operator curve and comparisons made between the performance of GMM and KLT/MI. Initial analysis was completed on the large limb muscles, MG and BB. The GMM had better accuracy than the KLT/MI when compared against operator identifications as the ground truth signal (88 – 94 % vs. 82 – 90 %). When EMG was used as the ground truth the GMM again had higher accuracy (81 – 88 % vs. 70 – 79 This thesis has shown a GMM computational analysis can detect fasciculations across a wide range of muscles and also can be used for the characterisation of fasciculations as they appear in ultrasound images, with significant differences being found between the healthy and MND affected participant groups. It has provided a foundation from which to build, with suggestions for future work being collecting images of stimulated twitches in a wide range of muscles for further characterisation and also a larger scale study prior to an official diagnosis being made to determine sensitivity and specificity values for this method as a diagnostic test.

Impact and Reach

Statistics

Downloads
Activity Overview
42Downloads
107Hits

Additional statistics for this dataset are available via IRStats2.

Actions (login required)

Edit Item Edit Item