Effect of Co-Occurrence Filtering for Recognizing Abnormality from Breast Thermograms

Nowadays, breast cancer is common and is on rise in the urban and even rural areas in India. Breast cancer is the massive growth of unwanted cells in the breast. Most women are affected by breast cancer, rarely in case of men. It causes the people to be more weak which makes them inactive physically and if left carelessly can even cause death. So it is important to observe for any abnormality in the breast in an earlier stage to avoid any complications in health. Fortunately, recent advances in machine learning has opened a new vista: Remote breast health monitoring using imaging modalities. Commonly used modalities are exposed to radiation. Infrared Imaging modality offers a solution to this challenge as this is a portable low cost non-invasive pain-free and radiation-free technique. In this work, we propose development of a Computer Aided Diagnostic (CAD) system for the recognition of abnormality in breast infrared images. Firstly, a co-occurrence filter based on edge-preserved technique is applied to preprocess the breast thermogram. Secondly, the Region of Interest (ROI) is extracted by adopting image cropping. Finally, relevant features are extracted from the ROI and fed to classifiers for recognizing abnormality in the breast thermogram.