Automated Segmentation for Quantitative Analysis of Breast Thermograms

Abstract

Thermography in breast cancer detection presents promising diagnosis when used as a first line screening tool in adjunct to mammography. A thermogram consists of various regions such as patient's arms, chest and breasts. Segmentation of the breast region is an important step in classification of normal and abnormal thermogram. This paper deals with image pre-processing using inpaint algorithm and automated segmentation of breast regions using anisotropic diffusion filter along with level-set evolution without reinitialization. Three types of breast thermograms which include one normal and two abnormal were considered for this study. The segmented results were tested using performance measures and 98% of accuracy is obtained.

Detection of Breast Cancer Using Infrared Thermal Images for Improved Accuracy by Using Random Forest and Multilayer Perceptron

“…Using thermography as an example, the authors of (Sathish et al 2016) surveyed the state of the art in computer-aided cancer detection techniques across a variety of medical imaging modalities. In (Selle et al 2015), the authors introduced a novel approach to ROI extraction from breast thermograms, one that considers both the breasts' lateral and frontal perspectives. The discovered ROIs assist doctors in distinguishing between normal and aberrant biomarkers.…”