Objective: The use of computer-aided diagnosis (CAD) methods makes significant contributions to the accurate and early diagnosis of breast cancer. For the segmentation step, which is one of the steps of CAD methods, there are different algorithms for image segmentation in the literature. Segmentation in detecting breast lesions using CAD methods may affect the features to be obtained from the images and accordingly the classification results. These segmentation methods have advantages and limitations compared to each other. In the literature, no study has been found in which feature matrices obtained by different segmentation methods are examined with different performance criteria. The aim of the study is to investigate the effects of different segmentation methods used in image processing on mammography images on breast cancer detection.
Materials and Methods: In the pre-processing step, the images are improved using median filtering, CLAHE and un-sharp masking. The texture features are extracted using texture analysis techniques from the region of interests (ROI) images and elastic network technique for feature reduction.
Results: The results obtained with five different segmentation algorithms were compared using performance measures (accuracy, sensitivity, specificity etc.) taking into account different classification methods. The k-means algorithm showed higher performance than other segmentation methods. The k-means algorithm exhibits good efficacy, accuracy of 1.00 and 0.989 were obtained in Support vector machine (SVM), Random Forest (RF) etc. classifiers, respectively.
Conclusion: As a result, segmentation methods used in image processing were found to have an impact on classification. These computer-aided systems can be used for patient classification.