On the expansion of training datasets to improve deep learning-based cell organelle recognition in fluorescence microscopy images of virus-infected cells

The Institute for Virology, Philipps-University, Marburg, developed a method to create image sequences of Marburg virus-infected live-cells. This work focuses on the expansion of image datasets by various transformation techniques to improve the training of a neural network for pattern recognition, i.e., the detection and classification of cell structures by means of pure green fluorescent imaging of subviral particles The results show a high potential for automated segmentation of cell organelles