Automated identification of breast cancer using digitalized mammogram images
Chachalo Gómez, Bryan Patricio · 2021
Cancer affects any organ uncontrollably invading and spreading along the body. According to World Health Organization breast cancer is on top of the leading cancers in affecting women around the world. Early treatment of people who develop cancer improves the prognosis of this disease and even saves lives. Unquestionably, in cancer diagnosis, the proper classification of carcinomas into benign, malignant and normal is a complex undertaking. An algorithm based on Computer Aided Diagnosis (CAD) is presented to detect breast cancer using mammograms. In this CAD implementation, transformations such as binarization, threshold smoothing and the main operation, Gabor wavelet, are used for preprocessing to suppress unnecessary labels and information and to obtain the best identifying features. We use techniques such as principal component analysis (PCA), t-distributed stochastic neighbor embedding (TSNE), and a collection of statistical variance models to identify features and reduce the feature space. Finally, we examine the k-Nearest Neighbors technique for classification (k-NN).