Nuclei spotting using deep learning

Abstract

Identification and classification of nuclei from microscopy is vital to new pharmaceutical developments. Biologist lacks a robust and efficient way to detect nuclei to natural variation in their appearances as well as differences in image capturing methods. Identification and classification of nuclei from microscopy images is considered as a complex task. A successful implementation will aid researchers immensely in their fight to find pharmaceutical solutions to medical crises while saving both valuable research time and funding. In this study, we employed a modified U-Net a deep learning based approach for nuclei detection where we computed 0.78 value of IOU (intersection over union) over validation set. This thesis was inspired by the Kaggle 2018 Data Science Bowl.