Efficient Classifier Training to Minimize False Merges in Electron Microscopy Segmentation

Abstract

The prospect of neural reconstruction from Electron Microscopy (EM) images has been elucidated by the automatic segmentation algorithms. Although segmentation algorithms eliminate the necessity of tracing the neurons by hand, significant manual effort is still essential for correcting the mistakes they make. A considerable amount of human labor is also required for annotating groundtruth volumes for training the classifiers of a segmentation framework. It is critically important to diminish the dependence on human interaction in the overall reconstruction system. This study proposes a novel classifier training algorithm for EM segmentation aimed to reduce the amount of manual effort demanded by the groundtruth annotation and error refinement tasks. Instead of using an exhaustive pixel level groundtruth, an active learning algorithm is proposed for sparse labeling of pixel and boundaries of superpixels. Because over-segmentation errors are in general more tolerable and easier to correct than the under-segmentation errors, our algorithm is designed to prioritize minimization of false-merges over false-split mistakes. Our experiments on both 2D and 3D data suggest that the proposed method yields segmentation outputs that are more amenable to neural reconstruction than those of existing methods.

ICON: An interactive approach to train deep neural networks for segmentation of neuronal structures

“…Compared to a CNN of the same architecture trained on all ground-truth images, our technique achieves slightly better segmentation accuracy. Our results corroborate well with past studies [3,7] that found that interactive training tends to produce a classifier that performs better for segmentation.…”

“…[3] and Parag et.al. [7] also proposed methods for sparse selection of a subset of training examples and demonstrated their advantages for EM segmentation. However, the strong dependence of the random forest classifiers on hand-tuned features has the potential to limit their performances on images from different EM preparation/imaging techniques and, more generally, from other data modalities.…”

FusionNet: A Deep Fully Residual Convolutional Neural Network for Image Segmentation in Connectomics

“…Instead of simultaneously considering all surrounding pixels and computing responses for the feature maps, RNN-based networks treat the pixels as a list or sequence with various routing rules and recurrently update each feature pixel. In fact, RNN-based membrane segmentation approaches are crucial for connected component labeling steps that can resolve false splits and merges during the post-processing of probability maps (Ensafi et al, 2014;Parag et al, 2015).…”

Detecting Synapse Location and Connectivity by Signed Proximity Estimation and Pruning with Deep Nets

“…Discovering the anatomical structure entails a 3D segmentation of EM volume. Numerous studies have addressed this task with many different approaches, we refer interested readers to [3] [4][5] [6] [7][8] [9] for further details. In order to unveil the connectivity, it is necessary to identify the locations and the direction of synaptic communications between two or more cells.…”