Özet:

Dokusal properties are widely used in the classification of mammographic images. This study aims to study the effects of the first and second degree tissue properties in the classification of mammography images. For this purpose, a computer-supported diagnosis (BDT) system was first designed. The computer-backed diagnosis system mainly aims to increase and support the accuracy of the diagnosis. The designed system consists of three steps as pre-processing, character extraction and classification. In the processing step, noise was first destroyed. Then the pectoral muscle was cleaned and the image improvement was carried out. The first and second degrees of tissue characteristics are calculated from improved mammogram images at the extraction stage. In the latter stage, mammography images are classified as normal and abnormal by the k-EYK algorithm, support vector machines (DVM) and linear discriminatory analysis (LDA) methods. From the point of view of the classification, it appears that the DVM method shows better performance than the k-EYK and LDA classifiers. Mammography classification was performed with 77.63% accuracy for LDA, 82.89% accuracy for k-EYK and 85.53% accuracy for DVM. The classification results obtained for each character group and classifier are compared. It has been observed that second-degree tissue properties increased the classification success.

Anahtar Kelimeler: