Quantitative Analysis and Prediction of Motor Neuron Morphology in Zebrafish Larvae

Neuron morphology is linked to the development and function of the nervous system. Zebrafish larvae are convenient intact vertebrate models for the visualization of neurons, particularly motoneurons and are therefore suitable for quantitative analysis of their structure throughout growth. However, reliable measurements of neuronal features from 3D microscopy stacks of the entire zebrafish are hard to obtain due to compact somata, thin axonal branches, and bright spinal cord-region signal varying strongly in shape, intensity, and spatial organization. A current bottleneck is also the reliance on manual tracing and labelling, which does not scale well with large datasets. The aim of this thesis is to develop a more automated workflow for the analysis of these fluorescent motor neurons in zebrafish larvae. The segmentation component focuses on separating biologically meaningful structures to extract quantitative features such as soma number, size, and volume, as well as axon length and branching-related measurements. In parallel, the thesis investigates brightfield-to-fluorescence prediction using machine learning, evaluating whether fluorescent signal can be inferred from brightfield microscopy. By applying morphological analysis to both real and predicted fluorescence images, it is possible to assess the ability of fluorescence prediction to capture biologically relevant signal and variation, supporting scalable image-based phenotyping.