Classification of Cell Nuclei Using CNN Features

This paper proposes a method to discriminate benignor malignantof cell image. Proposed method uses a CNN to extract features from the nuclei images detected from the original cell images and benignor malignantis automatically classified by two classes classificationwith SVM. This paper treats a cancer classification method using features obtained by using CNN from cell nuclei extracted from images of melanoma. The effectiveness of the proposed method has been confirmed by experiments from the viewpoint of cell classification of benignor malignant. It is shown that the difference between the number of cell nuclei diagnosed as normal and the number of cell nuclei diagnosed as cancer is clear in the benignand metastasis, while the difference in malignantis vague. Moreover, it is confirmed that N/C ratio of malignancyand metastasis was slightly higher than that of benign.