Classification of breast cancer tumors using feature selection and CNN

Breast cancer is one of the most dangerous diseases in the world and almost two million new cases are diagnosed every year. It starts from the breasts tissue and then spreads to other parts of the body. Early detection of breast cancer is important to save the life of a woman as it is related with a risen number of available treatment options. Benign and malignant are the major types of tumors and they are cancerous and non-cancerous, respectively. Benign is not dangerous since it does not destroy the nearby tissues and cannot spread or grow. Malignant tumor invades neighbouring tissues, blood vessels and spreads to other parts of the body by metastasis. Therefore, differentiating malignant from benign will help to detect breast cancer in its early stage. Nowadays, machine learning techniques are used to classify the tumor types hence the quality of lift is increased. Several years ago, there were so many breast cancer detection approaches proposed. These approaches are proposed by using one of the two types of dataset available such as medical imaging data and feature distribution data. In imaging data, the tumor portion is cropped and then detects whether it is cancer or not. In feature distribution data, multivariate attributes were taken from the digitized image of a Fine Needle Aspirate (FNA) of a breast mass to detect the cancer tumor. Those attributes are describe the characteristics of the cell nuclei present in the digitized image. Medical imaging related research requires more time and medical knowledge, therefore many authors chose the feature distribution dataset[11] to their research. The remaining parts of this paper are assigned as follows. Section 2 gives Related work of this research. Section 3 describes the methodology of this research such as pre-processing, classification model and performance evaluation criteria. Section 4 gives experimental setup as the details about the dataset and the experimental results. Finally, Section 5 gives the conclusion of this research