Texture analysis of jaw bone CT images for characterization of bone quality

Abstract

The objective of this work is to design and implement classifier framework to assist the surgeon for preoperative assessment of bone quality from Dental Computed Tomography images. Oral implants are gaining popularity and wide acceptance as many studies on oral implant treatment have revealed successful outcomes. The proposed work is subdivided into three main parts. They are the gray level texture analysis, multi-scale texture analysis and the classification system. In the first part of the work the performance of five popular statistical texture characterization methods are addressed. The first part gives a formal (statistical) definition of quottexturequot and a framework is described to characterize bone texture using gray level methods. Hence in the second part special attention is given to wavelets as a means for extracting texture features and the different types of separable wavelet transforms are discussed. This thesis examines several possibilities for feature extraction and clustering steps in particular, novel feature extraction and clustering schemes are introduced and compared to other known techniques. In the third part of the work Insertion torque values recorded at the time of surgery are used to subjectively evaluate the bone quality at different locations of the jaw. Univariate linear regression analysis was performed between texture parameters versus age and insertion torque using Pearson correlation coefficients according to sample distribution. In this thesis, texture analysis has been shown to detect morphological differences in the cancellous bone between various bone quality groups. This data has provided new insights into the mechanisms of change to cancellous bone structure with respect to age, gender and disease. The results of texture analysis are explained in the context of conventional bone histomorphometry and a priori knowledge to advance the understanding of cancellous bone architecture. newline newline newline