Research Paper
Year: 2020 | Month: October | Volume: 7 | Issue: 10 | Pages: 4-10
Automatic Breast Thermography Segmentation Based on Fully Convolutional Neural Networks
Mazhar B. Tayel1, Azza M. Elbagoury2
1Faculty of Engineering, Alexandria University, Alexandria, Egypt;
2Department of Basic Science, Faculty of Engineering, Pharos University, Alexandria, Egypt
Corresponding Author: Azza M. Elbagouryr
ABSTRACT
Breast cancer is one of the most cancer incidence the women world-wide. Early detection of breast cancer can increase the survival rate. Breast infrared thermography is a novel technology used for detecting early-stage breast abnormalities. The manual analysis of a huge number of these images is error-prone and consuming time and effort. Furthermore, the absence of clear edges and low contrast in thermogram may cause difficulty to analyze the image. Automated analysis of thermography images using machine learning increases the accuracy of detecting breast cancer and enables use in breast cancer screening programmes. A crucial step in the automated analysis is a segmentation, which is the aim of this paper.
The recent advances in deep learning, especially convolutional neural networks (CNNs), are making them the state-of-the-art methodology for automated image analysis. This paper presents an automated segmentation technique to extract the ROI for breast infrared thermography images, based on the use of Fully Convolutional Networks (FCN).
The experimental results prove that the accuracy, sensitivity, specificity of FCN reached to 96.4%, 97.5%, and 97.8% respectively. That denotes that the proposed automatic segmentation technique is an appropriate technique for extracting the breast ROI image from breast thermograms. Adequate comparison among recently segmentation technique by using available databases is very important, here some comparisons with other techniques are made. The comparison proves the superiority of using deep FCN over conventional algorithms.
Keywords: Breast Cancer, Thermography, Region of Interest, Segmentation, Deep convolutional neural network, deep fully convolutional networks.
[PDF Full Text]