Detecção de câncer de mama utilizando imagens termográficas

Abstract

Breast cancer is the leading cause of death in women worldwide. Early diagnosis is fundamental for a greater chance of recovery and healing of patients. The thermographic images due to the ability of early detecting the disease in young women, absence of radiation and the low cost has shown to be a promising technique. This thesis presents a methodology that classifies patients as normal, with benign and cancerous tumors (cancer) from thermographic images of the breast. The database used is composed of images of 70 patients. Some of the characteristics considered are statistical measures and fractal dimension. The classification is done through Artificial Neural Networks (ANN) and Support Vector Machines (SVM). The best result was 80.95% accuracy, 83.33% specificity for normal patients, 85.71% specificity for patients with benign alterations and 75% sensitivity. This result was obtained considering the characteristics: mean, standard deviation, median, minimum temperature, maximum temperature, thermal amplitude, skewness, kurtosis, entropy, contrast, correlation, moment 2, moment 3 and moment 4 to describe the images and using the classifier SVM. The results obtained are promising and confirmed the premise that thermographic images can be used in the detection of breast cancer.