M.Sc. Tezi Görüntüleme

      Todays, contribution to the development of the medical fields are achieving with medical sciences as well as engineering sciences. For diagnosis and treatment of diseases, medical hardwares and softwares which guides consultants are becoming increasingly important. The development of the medical imagining systems parallel to the technology have revealed the fact that these systems must be supported in software. To securely store the digital medical images and make them swiftly accessible can be handled with the help of the software systems which are in progress of development.

      This study includes the impelentation of thecniques that used obtain the lung tissue by segmentation of the chest CT images and the comperatively consideration of their success. Used methods are k-means and expectation-maximization clustering which are less common in image segmantation but more in classification. At the first stage, the image database have been created and then recorded images are manually segmented to calculate result images. In second stage, semi-automatic image segmantation carried out for the images by applying the preferred methods. By using the morphological operators small areas and non-lung tissue regions removed form the lastest images. Both method results for the accuracy, sensitivity and specificity values were calculated. In lung segmentation EM clustering method produced better results than the k-means clustering method.

      Key Words: Lung Segmentation, K-means Clustering, Expectation-Maximization, Fuzzy C-Means, Computer Aided Diagnosis