Patch-based All Convolutional Neural Network Model for Classification of Benign and Malignant Mammograms
Gurpreet Kaur, Annie Julie Joseph, P. N. Pournami · 2022 IEEE International Conference on Signal Processing, Informatics, Communication and Energy Systems (SPICES) · 2022
Breast cancer is one of the deadly diseases affecting middle-aged women. Early detection and accurate diagnosis are needed to improve the survival rate of the patients. Digital mammograms are one of the most frequently used techniques for helping early breast cancer detection. Radiologists face challenges for correctly diagnosing breast cancer as they have to examine many mammogram images daily. A computer-aided diagnosis (CAD)system is required to detect abnormalities as early as possible. The present study developed a deep learning model to differentiate malignant from benign tissues using mammograms collected from MIAS, INBreast, and DDSM datasets. As these images contain noise, Contrast limited adaptive histogram equalization (CLAHE) is used to preprocess the images. Then, mammographic segmentation for removing unnecessary background from a mammogram image is carried out. Then, these images passed to Multiscale All Convolution Neural Network(MACNN) to classify a mammogram into benign or malignant. This model has achieved a classification accuracy of 81%.In this study, improvement of the MACNN model, a patch-based MACNN, the proposed model learns more features than MACNN and has earned a classification accuracy of 88%.This model has better classification accuracy than other models ResNet50 and U-net model.