Our goal is to analyze neuronal morphology during development using artificially created templates. Such templates serve as input from the domain expertise: a standardized representation of the neuron, independent from imaging modalities and resolution, is what a neurobiologist can provide from knowledge and qualitative observations in the microscope. A template is divided into context-specific (topology, shape, and/or function) compartments, and our task is to segment input neuron volumes from their surroundings and partition them accordingly. We solve this problem using a global-to-local approach. We first employ a global transformation that serves as part-wise alignment of the template with the input volume. Then, we use local, deformable compartment shape registration using MRF-based free-form deformations. We validated our results using aCC motorneuron image stacks from larva Drosophila, at multiple developmental instances and different spatial resolutions.