3D-aware neural network for analyzing neuron morphology

With the continuous development of neuron reconstruction techniques, biologists can acquire neuron morphology data in large quantities. However, the vastness and richness of the data pose significant challenges for quantification and analysis tasks. In this paper, we propose an end-to-end 3D-aware neuron morphology analyzing model to extract morphological features. This method is able to capture the 3D relationships present in neuronal morphology data, thereby extracting cell-type-discriminative morphological features through an end-to-end pipeline. We tested our model with over 1,600 neuron morphology data collected from NeuroMorpho on classification and retrieval tasks. Our method is effective in both classifying neuronal cell types and retrieving neurons with similar morphology.