A quantitative census of millions of postsynaptic structures in a large electron microscopy volume of mouse visual cortex

Neurons display remarkable sub-cellular specificity in their synaptic targeting, which varies by cell type--for example, excitatory neurons prefer to target the spines of other excitatory cells. Modern dense neuroanatomy data, such as large volumetric electron microscopy connectomes, enable the study of this sub-cellular specificity and its context in a circuit at unprecedented scale and resolution. However, this scale has also made it challenging to create accurate and efficient methods for classifying and segmenting fine cell components (including spines) across entire volumes. Here, we present a cost-efficient computational pipeline for classifying postsynaptic targets and segmenting structures such as spines. Our method relies only on having a mesh representation of a neuron and avoids processing imaging data directly. Instead, we leverage tools from geometry processing to create features from the intrinsic geometry of a neurons surface. We couple this core technique with strategies for accelerating the computation and reducing the storage size of these features, creating a pipeline which can be deployed reliably over hundreds of thousands of neurons in the commercial cloud for a few hundred dollars. We then show how a simple but accurate classifier can use these mesh-based features to classify synapses as targeting somas, dendritic shafts, or spines (weighted F1 score 0.961). Using this pipeline, we create a publicly available map of the postsynaptic structures at over 208.6 million synapses in the MICrONS mouse visual cortex dataset. We present an overview of this census of postsynaptic targeting in MICrONS, finding expected patterns (e.g., excitatory neurons preferentially targeting excitatory spines) as well as less characterized exceptions (e.g., Layer 5 near-projecting and Layer 6 corticothalamic cells often connecting to excitatory neuron shafts). These tools also enable us to detect spines which receive multiple synaptic inputs--we find that the frequency of these multiply-innervated spines is unexpectedly variable across cells even within a cell type. We make our postsynaptic target predictions available for study, as well as the code for the computational pipeline and commercial cloud deployment. More generally, our work demonstrates how representations derived from neuronal meshes can be a powerful and scalable primitive for describing neural morphologies.