Espina: A Tool for the Automated Segmentation and Counting of Synapses in Large Stacks of Electron Microscopy Images

Abstract

The synapses in the cerebral cortex can be classified into two main types, Gray's type I and type II, which correspond to asymmetric (mostly glutamatergic excitatory) and symmetric (inhibitory GABAergic) synapses, respectively. Hence, the quantification and identification of their different types and the proportions in which they are found, is extraordinarily important in terms of brain function. The ideal approach to calculate the number of synapses per unit volume is to analyze 3D samples reconstructed from serial sections. However, obtaining serial sections by transmission electron microscopy is an extremely time consuming and technically demanding task. Using focused ion beam/scanning electron microscope microscopy, we recently showed that virtually all synapses can be accurately identified as asymmetric or symmetric synapses when they are visualized, reconstructed, and quantified from large 3D tissue samples obtained in an automated manner. Nevertheless, the analysis, segmentation, and quantification of synapses is still a labor intensive procedure. Thus, novel solutions are currently necessary to deal with the large volume of data that is being generated by automated 3D electron microscopy. Accordingly, we have developed ESPINA, a software tool that performs the automated segmentation and counting of synapses in a reconstructed 3D volume of the cerebral cortex, and that greatly facilitates and accelerates these processes.

Characterization and extraction of the synaptic apposition surface for synaptic geometry analysis

“…Stacks of serial images are then obtained by automated FIB/SEM (Merchán-Pérez et al, 2009). Since image segmentation, quantification, and analysis of synaptic junctions in these stacks are all labor-intensive procedures, we have developed ESPINA, a software tool that greatly facilitates and accelerates these processes (Morales et al, 2011a). ESPINA makes use of the fact that presynaptic and postsynaptic densities appear as dark, electron-dense structures under the electron microscope.…”

“…In the present study this was achieved with ESPINA, a software tool designed for the segmentation of electron microscopy images that uses heuristics based on gray levels and connectivity (Morales et al, 2011a). The result of the segmentation of a synaptic junction is a set of connected voxels that have a gray level over a certain threshold.…”

“…Therefore, if the segmentation of a given synaptic junction is defective, the SAS will also be defective. This can be avoided during the segmentation step by applying adequate gray-threshold settings and by eliminating noisy pixels in the original stack of images using a Gaussian blur filter, as described elsewhere (Morales et al, 2011a). …”

A Comparative Study of Feature Descriptors for Mitochondria and Synapse Segmentation

“…In [18], texture and shape-based features are extracted from small image patches around potential objects of interest. The approach used shows high accuracy and outperforms the features based on the correlation of pixels in [21] and the gray levels and connectivity features with automatic techniques designed to detect synapses in [16].…”

Domain Adaptation for Microscopy Imaging

“…1. Typically, a combination of manual and semi-automated segmentation or annotation tools such as [1], [2], [3] are then used to extract structures of interest. However, while the ever growing amount of available imagery should help unlock the secrets of neural functioning, the required amount of human annotation effort remains a major bottleneck.…”