Cyber Times International Journal of Technology & Management

A mammogram image segmentation and compression technique is proposed for classifying and storing information about breast cancer tissue. Initially a preprocessing is done in the mammogram images with Contrast limited adaptive histogram equalization (CLAHE). The features are extracted from the images. Then improved watershed transform is applied to the images for segmentation. Genetic training is applied to the images. Neural network classifier (genetic algorithm) is used for classification of breast cancer tissue in normal, benign, and malignant. Derivative based feature saliency technique is applied to the images for feature extraction. Vector quantization (VQ) with artificial neural network is used in the three regions with different compression rate based on the importance of the region and by this we also retain important tumor characteristics and minimize the size of mammogram images and thus reducing the storage space and increase its efficiency. At least the images undergo morphological process and micro classification area detection. This is specially done for benign and malignant regions. Two types of compression namely lossy and lossless compression are done in the segments based on the importance of those regions. Results show that this method gives accurate results in diagnosis of breast cancer and gives efficient result in the compression applications. The proposed method gives accurate results in differentiating malignant and benign tumor. It reduces the storage space when compared with other existing compression mechanisms. Thus the proposed system is more accurate in detecting tumor tissue and efficient in storing the images.