Project Summary: Electron Microscopy (EM) connectomics aims to build comprehensive maps of connections within an organism's nervous system. Serial EM provides us with detailed anatomical information about neurons and their connections.

Automated methods are needed to speed up the process of segmenting the neurons and reconstructing their 3D shape. In particular, Convolutional Neural Networks (CNN) are deployed to distinguish all the neurons and predict synapses locations.  This project focuses on building a framework to perform an analysis workflow for neural networks of the cerebellar structure, highlighting the role of morphologies, whose functions is still unclear.

The EM data consists of lobule V and X of the mouse cerebellum. What lobule V is known for is general sensory motor control with clear somatotopic organization, lobule X is mainly involved in vestibular and visual functions and it is evolutionary older. The neural networks in these two areas can be compared in terms of their connectivity. To achieve this goal, graph theory plays a crucial role in the analysis: neurons are modeled as nodes and their synapses represent the edges. The built graph can answer several questions related to the over- and underrepresented motifs in the connectivity, but it can also give more more insights concerning the probability of connection.

Eventually, machine learning and deep learning approaches will be explored to better classify the different cell types.