Use this URL to cite or link to this record in EThOS:
Title: Thickness estimation, automated classification and novelty detection in ultrasound images of the plantar fascia tissues
Author: Boussouar, A.
ISNI:       0000 0004 7960 2628
Awarding Body: University of Salford
Current Institution: University of Salford
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Access from Institution:
The plantar fascia (PF) tissue plays an important role in the movement and the stability of the foot during walking and running. Thus it is possible for the overuse and the associated medical problems to cause injuries and some severe common diseases. Ultrasound (US) imaging offers significant potential in diagnosis of PF injuries and monitoring treatments. Despite the advantages of US, the generated PF images are difficult to interpret during medical assessment. This is partly due to the size and position of the PF in relation to the adjacent tissues. This limits the use of US in clinical practice and therefore impacts on patient services for what is a common problem and a major cause of foot pain and discomfort. It is therefore a requirement to devise an automated system that allows better and easier interpretation of PF US images during diagnosis. This study is concerned with developing a computer-based system using a combination of medical image processing techniques whereby different PF US images can be visually improved, segmented, analysed and classified as normal or abnormal, so as to provide more information to the doctors and the clinical treatment department for early diagnosis and the detection of the PF associated medical problems. More specifically, this study is required to investigate the possibility of a proposed model for localizing and estimating the PF thickness a cross three different sections (rearfoot, midfoot and forefoot) using a supervised ANN segmentation technique. The segmentation method uses RBF artificial neural network module in order to classify small overlapping patches into PF and non-PF tissue. Feature selection technique was performed as a post-processing step for feature extraction to reduce the number of the extracted features. Then the trained RBF-ANN is used to segment the desired PF region. The PF thickness was calculated using two different methods: distance transformation and a proposed area-length calculation algorithm. Additionally, different machine learning approaches were investigated and applied to the segmented PF region in order to distinguish between symptomatic and asymptomatic PF subjects using the best normalized and selected feature set. This aims to facilitate the characterization and the classification of the PF area for the identification of patients with inferior heel pain at risk of plantar fasciitis. Finally, a novelty detection framework for detecting the symptomatic PF samples (with plantar fasciitis disorder) using only asymptomatic samples is proposed. This model implies the following: feature analysis, building a normality model by training the one-class SVDD classifier using only asymptomatic PF training datasets, and computing novelty scores using the trained SVDD classifier, training and testing asymptomatic datasets, and testing symptomatic datasets of the PF dataset. The performance evaluation results showed that the proposed approaches used in this study obtained favourable results compared to other methods reported in the literature.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available