Quantitative image analysis of peripheral nerves in whiplash injury patients

Anantharaman, Kamakshi Pradeep (2018) Quantitative image analysis of peripheral nerves in whiplash injury patients. Doctoral thesis (PhD), University of Sussex.

[img] PDF - Published Version
Download (3MB)

Abstract

The research in this thesis has examined the use of texture and shape analysis to characterise Magnetic Resonance (MR) images of peripheral nerves in order to provide a potential quantitative tool for better diagnosis and treatments.
Texture and shape can be considered as inherent properties of all surfaces and have the potential to provide sensitive information which cannot be quantitatively perceived by human vision. Texture analysis has been successfully used in image classification of aerial and satellite imagery and the diagnosis and prognosis of several types of cancer. However, to date, it has never been used in investigating peripheral nerve damage. In this thesis, we study the application of texture and shape analysis to the peripheral nerves in the upper extremities of patients suffering from Whiplash Associated Disorders (WAD).
Specifically, quantitative texture analysis was performed on MR images of the carpal tunnel which contains the median nerve. The median nerve was studied to identify differences in textural patterns. Texture methods such as: first order features; co-occurrence matrices; run-length matrices and autocorrelation function were applied and their performance was assessed. Texture analysis was also performed to investigate nerve damage in the MR images of the brachial plexus, both in controls and patients.
Further, spatial domain shape metrics were used to quantify and study the morphological differences of the median nerve in controls and patients. This highlighted that some significant differences exist between groups and thus could potentially be reliably used in combination with clinical scale metrics to identify possible nerve damage.
As MR images contain noise, locating the median nerve accurately to perform image analysis is very important. Therefore, we further investigated the application of an enhanced correlation filtering method that could be trained on images of the median nerve and then applied to detect the median nerve in test images. The Optimal Trade-off Maximum Average Correlation Height (OT-MACH) filter includes the expected distortions in the target in the construction of the filter reference function. The OT-MACH filter was tuned in a bandpass to maximize the correlation peak and thereby successfully locate the position of the median nerve in the carpal tunnel.
This study has successfully demonstrated that texture and shape analysis can be used to investigate possible peripheral nerve damage. Further research is required using larger datasets to establish a quantitative image analysis tool to support clinical decision making and thereby improve patient care and treatment outcome.

Item Type: Thesis (Doctoral)
Schools and Departments: School of Engineering and Informatics > Engineering and Design
Subjects: R Medicine > RZ Other systems of medicine > RZ0399 Osteo-magnetics, neuropathy, etc., A-Z
Depositing User: Library Cataloguing
Date Deposited: 17 Apr 2018 13:24
Last Modified: 17 Apr 2018 13:24
URI: http://sro.sussex.ac.uk/id/eprint/74969

View download statistics for this item

📧 Request an update