The primary objective of this thesis is to investigate the possibilities of different U-Net-like neural networks for segmenting tunneling nanotubes in multidimensional fluorescence microscopy image data of Drosophila's wings. The textual part of the thesis must include a concise and comprehensive description of the evaluated networks and their quantitative as well as qualitative performance comparison focused on segmentation accuracy, execution time, and memory consumption. The implementation of the evaluated networks, including training protocols and pre-trained models, will be delivered in an easy-to-use form and complemented with a practical guide that documents the exemplary usage of individual networks.