Automated detection of breast cancer in mammography images using Gabor filter and Bayesian classifier / Bardia Yousefi

The cause of breast cancer is unknown, also there are some prescriptions from doctors regarding less of stress and having good life habit, so early detection can reduce the death rate of patients. The earlier and accurate detection of symptom helps to provide re-liable diagnosis and better treatment. Some of researches in the area of detection of cancer in breast images have been done. The results which obtain within researches are significantly considerable whereas working on the area is one of the imperative researches in the area of biomedical engineering and researchers are still looking for the better feature extraction method for this problem. The problem of research in the area of diagnosis breast cancer via pattern recognition and image processing techniques introduced when the existed methods are have some limitations in detecting and classifying the breast lesions have not been accurate. The objective of this research is developing a technique for feature extraction in breast lesions by applying Gabor filter and using Bayesian classifier for classification of the breast lesions. The extraction of features using Gabor filter is a new technique for feature extraction regarding breast lesions in mammographic images. Moreover, the aim this application having an efficient features extraction technique and better classifying of lesions. This method focuses on developing a technique for feature extraction in breast lesions by applying Gabor filter and Bayesian classifier. First the Gabor filter will be applied to the breast image as feature extractor and then the extracted features will be used for Bayesian classifier. The results will represent breast lesions which indicate the breast cancer. The proposed method will be applied to 40 mammographic images out of which 10 cases are normal patients with no sign of breast cancer while 30 are breast cancer patients. We expect that the extracted features by using the proposed approach represent reliable results and significant improvement about 97.5 percent of accuracy in detection of breast lesions