A Hybrid Model for Mammogram Images Classification Using Transfer Learning
Kamal Nain Sharma, Amit Kamra · Zenodo (CERN European Organization for Nuclear Research) · 2021
Abnormality detection in a mammogram image is very important to save the life of a patient. Many automatic CAD systems are used for mammogram screening and classification, but still, accuracy is one of the major concerns. In this paper, a multistage mammogram image classification model has been proposed which can classify the mammogram images as mass and micro calcification. The proposed multistage model based on enhancement, segmentation, and classification training had shown an accuracy of around 99.15%. Transfer learning is used to train and test the network. In this work, publically available MIAS and DDSM datasets are used to acquire various mammogram images. The network is also tested on various real-time clinical images and the network accuracy is still near about 99%. The work done will add value to the existing literature as the proposed model can classify the images with much higher accuracy.