| Summary: In recent years Cone-beam Computed Tomography (CBCT) imaging is frequently used for radiological examinations in the field of dentistry. In this thesis, studies on segmentation and classification of lesions of periapical cyst (PC) and keratocystic odontogenic tumor (KOT) in three dimensional (3D) CBCT images have been carried out and methods have been proposed. 50 CBCT 3D image dataset files have been employed as the dataset of the study. Experts have identified half of the data as periapical cyst (PC) and the other half as keratocystic odontogenic tumor (KCOT). The diagnosed lesions were used in experiments with specially developed software for this study. The lesional volumetric regions were manually segmented with the developed software tools to obtain a ground truth set. A noise removal approach has been proposed for the preprocessing step for removing noise detected in the CBCT images. A semi-automated segmentation approach has been proposed to accommodate the anatomical and internal variations of the lesions. Basic statistics and 3D Gray Level Co-Occurrence Matrix (GLCM) information were calculated from the segmented lesions and feature vector containing 636 feature information was obtained. Six different classifiers were used for classification experiments. As a result of classification experiments, PC and KCOT lesions can be detected and classified with great accuracy.
Key Words: Computer aided diagnosis; Dental lesion; Cone beam computed tomography; Periapical cyst; Keratocystic odontogenic tumor; Volumetric textural features; Noise removal; Semi-automatic segmentation; Classifier; Dental image database | 
