A Nucleus Based Feature Extraction From Histopathology Images Using CNN For Liver Cancer
P. Sabitha, G. Meeragandhi · 2022
The process of evaluating a surgical tissue specimen is known as histopathology. For preservation, tissue is fixed and placed on glass slides with chemicals. Whole slide imaging WSI) is the process or scanning conventional glass slides in order to produce digital slides, WSI continues to gain attraction among pathologists for diagnostic, educational, and research purposes. A WSI contains more than 100,000 pixels in any spatial dimension. Feature extraction plays an important role in the diagnosis of cancer using computer-aided methods. Various machine learning methodologies is introduced to extract features from WSI images. These approaches ignore to extract the large scale architectural relationships present in the nucleus. Nucleus in the WSI contains enormous information. For histopathology images, a nucleus-based feature extraction utilising a Convolutional neural network is proposed in this study. Initially the nucleus is detected from images which is then used in training a neural network. This helps in extracting macro and micro image level information like the spatial distribution and pattern. The extracted features are valuated using the classification approaches on a histopathology database of liver tumour. The results show a better classification performance for liver tumour when comparing other methods.