1. UKM Learning and Research Repository
2. PHD Thesis / Tesis PHD
3. Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Title:	Multi-classifier approach to x-ray medical image annotation and retrieval
Authors:	Mohammed Muayad AbdulRazzaq (P45075)
Supervisor:	Shahrul Azman Mohd Noah, Prof. Dr.
Keywords:	Medical image X-ray Multi-classifier Retrieval Dissertations, Academic -- Malaysia
Issue Date:	27-Jun-2016
Description:	As digitized medical images have been continuously increasing for the last two decades, medical images form an essential information source for diseases, surgical planning, medical reference, research and training. Effective techniques for medical image classification and retrieval are needed to provide accurate and quick search through a large amount of images. The huge amount of medical images makes it impossible to manually annotate them with meaningful textual annotations. Consequently, researchers moved toward using image visual features for indexing and retrieval. However, high level of semantic concepts are required to retrieve medical images which are difficult to be attained by using low-level features alone, this is called semantic gap. To remedy these limitations, an automatic mechanism is needed to classify and annotate medical images in different semantic levels. Assigning keywords or concepts by computer systems automatically to medical images based on semantic concepts are referred to automatic image annotation, which can provide a platform to bridge the semantic gap. This research aims to formulate an effective automatic medical images annotation and retrieval system based on using multiclassifiers. The objectives of this research are: firstly, to propose suitable features extraction algorithm for X-ray medical images. Secondly, to formulate the semantic model for automatic medical image annotation and classification of the X-ray medical image. Thirdly, to identify the methods of machine learning for automatic image annotation of the X-ray medical image. Fourthly, to evaluate different image features (global, local, and combined) and classifiers in retrieving the X-ray medical image. This research proposes an approach for X-ray medical image annotation and retrieval based on using multi-classifiers. Multi-level feature extraction, multi classifications techniques, and concept hierarchy were used in the proposed approach. Pixel features were extracted in the first level of feature extraction. Using texture and shape features in the second level of feature extraction, texture features by using Gray Scale Cooccurrence Matrix (GLCM) and Wavelet Transform (WT), and shape features extracted global features by using the histogram of edge features. In the third level, local features were extracted based on dividing each image into four blocks then GLCM and WT from each block were extracted. In the fourth level, Speed Up Robust Feature (SURF) was applied. All these features were combined in one vector for each image, and the technique of Principle Component Analysis (PCA) was applied for feature reduction. Support Vector Machine (SVM) and k-Nearest Neighbour (k-NN) classifiers were used for annotation. Hierarchy concepts were design based on the body region, sub-body region and image orientation. ImageCLEF2005 database was used for evaluations. The obtained results in this research show improvement when compared with previous related studies.,Certification of Master's/Doctoral Thesis" is not available
Pages:	183
Publisher:	UKM, Bangi
Appears in Collections:	Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.