Evaluating Deep Learning Feature Towards Ranking of Immunocytochemistry Images
Pouya Ahmadvand, Behnam Maneshgar, Payam Ahmadvand · 2019
The analysis (figuring out that how well antibodies are linked to the Nucleus of interest) of Immunocytochemistry (ICC) images is mostly done manually. In this paper, we propose an automatic classification process to rank the images according to adherence of antibodies to the nucleus. To have a robust computerized diagnosis system, it is necessary to design a classifier with meaningful features. In this work, we evaluate two feature extractions based on deep neural network and engineered feature extraction. We proposed the first machining learning approach for ranking of Immunocytochemistry (ICC) images, while some of the automation works have been proposed for Immunohistochemistry (IHC) images. We tested our method on 300 microscopic images collected by ourselves and reached a reasonable classification ranking accuracy of 59.4% using all features.