Christopher Günther, Nickels Winkler, Jan N. Peters, Cora Lohse, Bernd Gromoll
Indivumed Services GmbH, Hamburg, Germany
The detection and segmentation of cell nuclei is a crucial step in most digital image analysis workflows. The accuracy of the recognition results is of utmost importance, especially when advancing to more sophisticated target recognition and quantification. As deep neural network algorithms are increasingly used to recognize nuclei of different shapes, sizes, and staining intensities, it is of interest to compare the detection result of a Deep Learning (DL) algorithm with a conventional threshold (TH) approach and to see how a pathologist would evaluate the same region of interest.
