Skip to the content

Automated speckle tracking in ultrasound images of tendon movements

Mohamed, ASA 2015, Automated speckle tracking in ultrasound images of tendon movements , PhD thesis, University of Salford.

[img]
Preview
PDF
Download (39MB) | Preview

Abstract

The central aim of this thesis was to develop new tracking software employing various image tracking algorithms for tracking the speckled movement of the tendon image captured using dynamic B-mode ultrasound imaging. The algorithms were selected based on the literature related to the tracking of images captured using ultrasound imaging. Experiments were carried out to validate these tracking algorithms in order to enable development of the tracking software. The experiments conducted paralleled the objectives in designing, developing, experimenting and implementing the image-tracking algorithm to track movement of the human tendon in vivo within the speckled ultrasound images. The development of the tracking software focuses on solving the problems of tracking the ultrasound images as well as analysing the tracking movement frame-by-frame to produce useful measurements that can be used to describe the localised mechanical and structural properties of the human tendon. The algorithms tested were Normalised Cross Correlation (NCC), Mean Square Error (MSE), optical flow – Lucas-Kanade (LK) and combination of NCC and MSE (NCCMSE) selected by signal-to-noise ratio (SNR) and were tested on both active and passive movements of the patella tendon (knee) and the medial gastrocnemius tendon (ankle). The comparison of the algorithms led to the identification of a single algorithm giving optimal result. The results from all tested algorithm showed NCC to be the closest match to the standard manual measurement. NCC was also the fastest among the algorithms tested and contained fewer errors in tracking. For NCC algorithm, various sizes of the region of interest (ROI) block were also tested and found that 15x15 pixels ROI block size gave the optimum measurement, which was close to the standard manual measurement. The threshold levels also indicated that >0.90 to be the optimum level for optimum tracking. The 2- ROI tracking analysis were also explored to look at the tracking performances when tracking at two different regional sites of the tendon simultaneously, and again the  NCC performed better with 15x15 ROI block size and comparable to the results obtained from the standard manual measurement. Lastly, multiple layers of the tendon were also explored to look at the excursion of the anterior, midsection and posterior layers of the tendon during ramped isometric contraction. This experiment uses all the settings found from previous experiment results, and applied to look at the mechanical properties of the human tendon. The experiments showed that the anterior gave the highest mean stain followed by the mid section and the smallest mean strain was found at the posterior proximal. The experiment also looked at the distal strain, with the result showing that the posterior gave the highest mean strain followed mid section and anterior layer gave the smallest mean strain. The experiment also looked at the performance of posterior layers and distal layers at 50 and 100% force levels. The experimental results showed that the NCC to be the optimum-tracking algorithm. The method described here has the potential to improve clinical knowledge relating to the tendon mechanical properties. The information generated by the tracking algorithm could help to give further insight into the aetiology of tendon injury, repair, response to various training interventions and the time course of tissue adaptation with disease.

Item Type: Thesis (PhD)
Contributors: Ritchings, RT (Supervisor) and Pearson, S (Supervisor)
Themes: Health and Wellbeing
Media, Digital Technology and the Creative Economy
Schools: Schools > School of Computing, Science and Engineering
Funders: Universiti Sains Malaysia
Depositing User: ASA Mohamed
Date Deposited: 09 Nov 2015 15:37
Last Modified: 05 Apr 2016 19:30
URI: http://usir.salford.ac.uk/id/eprint/35757

Actions (login required)

Edit record (repository staff only) Edit record (repository staff only)

Downloads

Downloads per month over past year