Deep Convolutional Neural Network System Based Classification of Digital Mammograms
P. Indra, Miss R.Yoganapriya · Research Square · 2023
Abstract Breast cancer is very common and considered as the second dangerous disease all over the world due to its mortality rate. So, if the detection is early enough, it can reduce the death rate. Image processing techniques are applied to accurately segment the Region of Interest (ROI) prior to abnormality detection in digital mammograms. The digital mammograms can majorly classify into two types, normal and abnormal. Abnormal cases are taken for further process. In this paper, some of the Non-linear techniques are applied to the mammogram images for the removal of noise at pre-processing. Noise removal is done by using lee filter, frost filter, median filter and improved statistical based bilateral filter. The best filter is selected by measures of MSE, PSNR and SSIM. Watershed, K-means clustering algorithms has been applied to the filtered image to get the accurate result prior to segmentation. For feature extraction, features from mammogram are calculated by using GLCM and Tetrolet Transform. Deep Convolutional Neural Network (DCNN) is used for classification of the images. Mammogram from MIAS and DDSM database is taken for simulation.