Classification of nuclei in whole slide imaging based on ensemble deep learning

Recently, several innovative digital imaging technologies led to the development of several medical imaging data representations. Whole slide imaging (WSI) refers to capturing and storing tissue glass slides and biopsy samples in a digital form. This newly emerging modality is increasingly gaining huge interest in pathology departments worldwide due to the unique advantages introduced by this medical imaging modality for diagnostic, educational, and research purposes. This work provides a comparative analysis of multiple classification approaches of nuclei in WSI where multiple nuclei context-aware manipulation strategies are evaluated by our nuclei classification experiments. We propose a contextaware deep ensemble approach based on ResNet deep architecture using different forms of input images pre-processed using multiple manual feature extraction techniques. Results show that our approach attains higher accuracy performance compared to traditional context-ware strategies in the literature.