Chapter 11 - Deep learning-based detection and classification of adenocarcinoma cell nuclei

Abstract

Nowadays, clinical practice uses digital pathology for examining digitized microscopic images to identify diseases like cancers. The main challenge in examining microscopic pictures is the need to dissect every single individual cell for precise analysis because identification of cancerous diseases depends emphatically on cell-level data. Due to this reason, the detection of cells is a significant point in medical image examination, and it is regularly the essential prerequisite for disease classification techniques. Cell detection and then classification is a problematic issue because of diverse heterogeneity in the characteristics of the cells. Deep learning methodologies have appeared to deliver empowering results on analyzing digital pathology pictures. This chapter introduces an approach by using region-based convolution neural networks for locating the cell nuclei. The region-based convolution neural network estimates the probability of a pixel belonging to the core of the cell nuclei. Pixels with maximum probability indicate the location of the core of the cell nucleus. After finding the cells, they are classified as healthy or malicious cells by training a deep convolution neural network. The proposed approach for cell detection and classification is tested with the adenocarcinoma dataset. Cell image analysis based on deep learning techniques shows good results in both the identification and classification of the cell nucleus.