Supervised Classification of Histopathological Images Using Convolutional Neuronal Networks for Gastric Cancer Detection

This paper presents a pilot-test for gastric cancer detection using histopathological image samples taken from pathology laboratory of the Universidad Industrial de Santander. The proposal follows two approaches: the first one analyzes the image as a whole to find characteristic properties of the benign samples, such as, the correct organization of the cells to form glands; the second one considers local morphological features, such as, the size, the shape, and the intensity of the light of individual nuclei in the cells. For both approaches, the automatic detection is based on a supervised classification model using a literature deep convolutional neural network. Experiments show a detection accuracy of up to 89,72%, which indicates that the proposal is a promising tool to assist the pathological diagnosis.