JCVirusNEntamoebaNET (JCENET): A Deep Neural Network for John Cunningham Virus and Entamoeba Parasite Detection

摘要

Parasites and viruses are responsible for attacking the immune system of human body and destroying it in the process which makes detection of parasites and virus a penultimate task in modern medical analysis. The detection process is often characterized with symptoms, microscopic tests. However, little research work has been carried out to exploit and garner the capability of deep learning to classify and detect the parasites and virus. In our research work, we propose a new deep neural network based image processing approach to classify the parasites, germs and virus for the case of Entamoeba parasite which leads to Amebiasis disease and John Cunningham virus which is responsible for Progressive multifocal leukoencephalopathy (PML). We have applied our own deep convolutional neural network to classify the JC virus and entamoeba using immunohistochemistry(IHC) images and microscope images from a dataset that We have collected from various neuropathology laboratories, and researchers. Our model achieved overall classification accuracy of 77% and F1-score of 76% with a strong narrative of rationale for utilization of batch normalization and dropout layer in deep learning model.