Self-Supervised Learning of Synapse Types from EM Images

TL;DR

Problem: labeling synapse types from EM images is valuable but requires costly ground-truth annotations. Approach: the authors propose a self-supervised contrastive learning framework using a VGG encoder trained on $80^3$-voxel EM crops from the Drosophila CNS, with ground-truth transmitter IDs used only for evaluation, and representations analyzed with $UMAP$. Findings: the learned representations separate synapses without explicit labels, and neurotransmitter ID does not appear as a primary driver in the main structure; the method can still initialize transmitter-prediction and assist automated proofreading to detect false neuron merges. Significance: this approach reduces annotation burden, offers a principled way to sample ground-truth across synapse structure, and may generalize across imaging conditions and species.

Abstract

Separating synapses into different classes based on their appearance in EM images has many applications in biology. Examples may include assigning a neurotransmitter to a particular class, or separating synapses whose strength can be modulated from those whose strength is fixed. Traditionally, this has been done in a supervised manner, giving the classification algorithm examples of the different classes. Here we instead separate synapses into classes based only on the observation that nearby synapses in the same neuron are likely more similar than synapses chosen randomly from different cells. We apply our methodology to data from {\it Drosophila}. Our approach has the advantage that the number of synapse types does not need to be known in advance. It may also provide a principled way to select ground-truth that spans the range of synapse structure.

Figures (2)

• Figure 1: Synapses classified by neurotransmitter type. Two dimensional UMAP reduction of classification output. Each neurotransmitter is given a distinct color.
• Figure 2: Results of self-supervised classification using only Dale's principal and segmentation.