Analysis of Methods of Classification of Breast Thermographic Images to Determine their Viability in the Early Breast Cancer Detection

Recent studies have shown that thermographic examination can be a very promising auxiliary tool in the early detection of breast cancer, a key factor in improving the patient's chances of cure. The present work aims to analyze several methods of classification of digital images, including infrared (IR) or thermographic images. It is intended to evaluate the results obtained with the objective of investigating the feasibility of the use of these images as an auxiliary exam for the detection of breast cancer. Initially IR images were acquired and processed. Then, the extraction of characteristics was performed, based on the appropriate temperature ranges from the thermograms. Thus, the input data were determined for the classification process. Several image classifiers were evaluated. Finally, 93.42% accuracy, 94.73% sensitivity and 92.10% specificity were the rates achieved for the Cancer Class in a binary (Cancer versus Non-Cancer) analysis. In a multiclass analysis (Malignant, Benign, Cyst and Normal), 63.46% accuracy, 80.77% sensitivity and 86.54% specificity were the rates achieved for the Malignant Class.