Please use this identifier to cite or link to this item:
Title: Classification of Texture Parameters Between Normal And Tumor Portions of a CT Scan Image
Authors: Kumar, Suchit
Supervisor: Mittal, Deepti
Keywords: SVM;GLCM;texture;classification
Issue Date: 19-Aug-2014
Abstract: Medical imaging has provided a non invasive way of extracting information from the internal organs of humans. Tumor diagnosis is one of the key factors for which medical imaging such as computer tomography CT scan are used. However medical images such as CT scan may not provide accurate information of the tumor or may be the information is based on human perception. Hence medical images have to be analysed in order to diagnose tumors accurately. Texture analysis of a CT scan image may help in differentiating between normal and the tumor portions. In this thesis we have used abdominal CT scan with tumor region marked on them. Texture parameter were calculated using the statistical approach proposed by haralick[1] called grey-level co-occurrence matrix(GLCM). Further a comparison was carried out between normal and the tumor portion texture parameters using box-plots. On the basis of the box-plot comparison further classification between normal and the tumor portion using values of texture parameters was performed .SVM classifier with different kernel functions was used for classification purpose. Further more we used a image enhancement technique called the adaptive histogram equalization technique to enhance the contrast of the CT scan images. The texture parameters calculated from the normal and the tumor portion of the enhanced images were subjected to the same procedure as the original images. This study investigates the texture features which are able to discriminate between normal and the tumor portions in a CT scan image both original and enhanced.
Description: M.E. (Electronic Instrumentation and Control)
Appears in Collections:Masters Theses@EIED

Files in This Item:
File Description SizeFormat 
2988.pdf2.21 MBAdobe PDFThumbnail

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