Fully-Automatic Synapse Prediction and Validation on a Large Data Set

Abstract

Extracting a connectome from an electron microscopy (EM) data set requires identification of neurons and determination of connections (synapses) between neurons. As manual extraction of this information is very time-consuming, there has been extensive research efforts to automatically segment the neurons to help guide and eventually replace manual tracing. Until recently, there has been comparatively little research on automatic detection of the actual synapses between neurons. This discrepancy can, in part, be attributed to several factors: obtaining neuronal shapes is a prerequisite for the first step in extracting a connectome, manual tracing is much more time-consuming than annotating synapses, and neuronal contact area can be used as a proxy for synapses in determining connections. However, recent research has demonstrated that contact area alone is not a sufficient predictor of a synaptic connection. Moreover, as segmentation improved, we observed that synapse annotation consumes a more significant fraction of overall reconstruction time (upwards of 50% of total effort). This ratio will only get worse as segmentation improves, gating the overall possible speed-up. Therefore, we address this problem by developing algorithms that automatically detect presynaptic neurons and their postsynaptic partners. In particular, presynaptic structures are detected using a U-Net convolutional neural network (CNN), and postsynaptic partners are detected using a multilayer perceptron (MLP) with features conditioned on the local segmentation. This work is novel because it requires minimal amount of training, leverages advances in image segmentation directly, and provides a complete solution for polyadic synapse detection. We further introduce novel metrics to evaluate our algorithm on connectomes of meaningful size. When applied to the output of our method on EM data from Drosphila, these metrics demonstrate that a completely automatic prediction can be used to effectively characterize most of the connectivity correctly.

En bloc preparation of Drosophila brains enables high-throughput FIB-SEM connectomics

“…Each offers particular advantages, but most important for our purposes, we report a method to improve the imaging speed of FIB-SEM by adopting novel ways to increase specimen contrast, and we apply these to an entire microdissected hot-knife sliced fly's brain comprising connected sub-and supraesophageal ganglia. Our methods are adapted to a FIB-SEM imaging mode and reliably recover fixed neurons as round cross-sections, suitable for machine segmentation (Parag et al, 2015), with dark synaptic profiles suitable for automated synapse detection (Huang et al, 2018). The numbers of the latter match closely the numbers of those identified by human proof-readers (Shinomiya et al, 2019) and so are considered accurate.…”

“…Synaptic sites ( Fig. 2b,d) could be clearly detected semi-automatically from their increased electron density (Huang et al, 2018), with typically a single T-shaped presynaptic density (or T-bar), opposite which sit a number of postsynaptic processes.…”

Automatic Detection of Synaptic Partners in a Whole-BrainDrosophilaEM Dataset

“…Recent neuron segmentation algorithms provide initial segmentations that significantly reduce the time needed to trace a Kenyon cell to about 8.6 s/µm or 201 days for all Kenyon cells. The relative time spent on manually identifying synaptic connections has thus increased significantly (Huang et al, 2018;Dorkenwald et al, 2017), highlighting the need for automatic methods for synapse annotation.…”

“…2a), whereas in the second step neuron segments (as induced by a neuron segmentation) in physical contact are classified as synaptically "connected" or "not-connected". For the first step, convolutional neural networks (CNNs) were successfully used to detect synaptic clefts (Heinrich et al, 2018) or T-bars (Huang et al, 2018) in large-scale fruit fly EM data. For the second step, Kreshuk et al (2015) proposed a graphical model to solve synaptic partner assignment given a neuron segmentation and candidate synapse detections.…”

“…More recently, CNNs have been used to infer synaptic connectivity: use a CNN to predict pre-and post-synaptic neuron masks from synaptic cleft segmentations and Parag et al (2018) use a CNN to classify neuron pair candidates extracted from a neuron segmentation as "connected" or "notconnected". Huang et al (2018) solve a similar classification task with a multilayer perceptron and local context features. In contrast to these two-step approaches, Buhmann et al (2018) jointly detect pre-and post-synaptic sites and infer their connectivity via longrange affinity edges using a single CNN architecture.…”