Detection of breast cancer in infrared images using Discrete Wavelet Transform and Support Vector Machine

This work presents a comparative study between two approaches which aim at helping diagnosing patients with breast cancer using thermal images. The first method uses the Discrete Wavelet Transform (DWT) for feature extraction and Support Vector Machine (SVM) to classify the patients. The other method is based on feature extraction with pretrained convolutional neural networks in conjunction with deep neural networks to perform the classification. VGG16, ResNet50 and MobileNet were considered to determine which was best suited to classify patient infrared images. Experimental results showed that DWT in conjunction with SVM presented the best performance on classifying 928 images (489 healthy and 429 sick) with 98% accuracy, 97% sensibility and 98% specificity.