Classification of Mammograms using Attention Learning for Localization of Malignancy
Manaswini Nagaraj1, Vignesh Prabhakar2, Sailaja Thota3
1Manaswini Nagaraj, Department of Computer Science Engineering, REVA University, Bangalore (Karnataka), India.
2Vignesh Prabhakar, Department of Computer Science Engineering, SRM Institute of Science and Technology, Chennai (Tamil Nadu), India.
3Sailaja Thota, Department of Computer Science Engineering, REVA University, Bangalore (Karnataka), India.
Manuscript received on 04 June 2019 | Revised Manuscript received on 12 June 2019 | Manuscript Published on 29 June 2019 | PP: 84-90 | Volume-8 Issue-5S, May 2019 | Retrieval Number: E10180585S19/19©BEIESP
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Mammography is a specialized medical imaging that uses a low-dose x-ray system to examine the breasts. A mammogram is a mammography exam report that helps in the detection and diagnosis of breast diseases in women at an early stage. This project proposes to classify mammography breast scans into their respective classes and uses attention learning to localize the specific pixels of malignancy using a heat map overlay. The attention learning model is a standard encoder-decoder circuit wherein convolutional neural networks perform the encoding and recurrent neural networks perform the decoding. Convolutional neural networks enable feature extraction from the mammography scans which is thereafter fed into a recurrent neural network that focuses on the region of malignancy based on the weights assigned to the extracted features over a series of iterations during which the weights are continuously adjusted owing to the feedback received from the previous iteration or epoch. Mammography images are equalized, enhanced and augmented before extracting the features and assigning weights to them as a part of the data preprocessing procedures. This procedure would essentially help in tumor localization in case of breast cancers.
Keywords: Attention Learning, Convolutional Neural Networks, Encoder-Decoder, Recurrent Neural Networks.
Scope of the Article: Classification