Multicut brings automated neurite segmentation closer to human performance

Beier, Thorsten; Pape, Constantin; Rahaman, Nasim; Prange, Timo; Berg, Stuart; Bock, Davi D.; Cardona, Albert; Knott, Graham; Plaza, Stephen M.; Scheffer, Louis K; Koethe, Ullrich; Kreshuk, Anna; Hamprecht, Fred A.

doi:10.1038/nmeth.4151

Cited by 145 publications

(163 citation statements)

References 2 publications

Supporting

Mentioning

161

Contrasting

Order By: Relevance

“…It might take the better part of a year merely to acquire the raw data (Schalek et al, 2016). While even a few years ago it seemed impossible to analyze such an amount of data within a reasonable time frame, recent progress in the automation of segmentation are encouraging (Berning et al, 2015; Januszewski et al, 2016; Beier et al, 2017; Dorkenwald et al, 2017). …”

Section: Discussionmentioning

confidence: 99%

EM connectomics reveals axonal target variation in a sequence-generating network

Kornfeld

Benezra

Narayanan

et al. 2017

eLife

108

View full text Add to dashboard Cite

The sequential activation of neurons has been observed in various areas of the brain, but in no case is the underlying network structure well understood. Here we examined the circuit anatomy of zebra finch HVC, a cortical region that generates sequences underlying the temporal progression of the song. We combined serial block-face electron microscopy with light microscopy to determine the cell types targeted by HVC(RA) neurons, which control song timing. Close to their soma, axons almost exclusively targeted inhibitory interneurons, consistent with what had been found with electrical recordings from pairs of cells. Conversely, far from the soma the targets were mostly other excitatory neurons, about half of these being other HVC(RA) cells. Both observations are consistent with the notion that the neural sequences that pace the song are generated by global synaptic chains in HVC embedded within local inhibitory networks.DOI: http://dx.doi.org/10.7554/eLife.24364.001

show abstract

Section: Discussionmentioning

confidence: 99%

EM connectomics reveals axonal target variation in a sequence-generating network

Kornfeld

Benezra

Narayanan

et al. 2017

eLife

108

View full text Add to dashboard Cite

show abstract

“…Even if we cannot identify the hierarchy class of a given architecture in a hierarchy, proving bounds [6] on them might give us some insight. Also, as alluded to before, these bounds can be used to establish that one set of networks is more complex than another.…”

Section: Definition 2 (Transformation Hierarchy) a Transformation Hiementioning

confidence: 99%

“…[5] The ordering is by set inclusion. [6] Again, formally, the bounds are with respect to the partial ordering induced by set inclusion. Bounds on hierarchy classes of specific sets of networks can be used to establish complexity results.…”

Section: Definition 2 (Transformation Hierarchy) a Transformation Hiementioning

confidence: 99%

“…Progress in experimental techniques, since the turn of the century [2,3,4,5,6,7], has led to a rekindling of interest in determining exact structures of (microscale) neural circuits that comprise nervous systems [8,9,10,11,12], which were rechristened as connectomes [13]. The first connectome -that of the nematode Caenorhabditis elegans -was determined in 1986 [14], after an effort that took over a decade, much of it manual.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Hierarchies & Lower Bounds in Theoretical Connectomics

Ramaswamy

2019

Preprint

View full text Add to dashboard Cite

Connectomics is a sub-field of Neuroscience aimed at determining connectomes -exact structures of neurons and their synaptic connections in nervous systems. A number of ongoing initiatives at the present time are working towards the goal of ascertaining the connectomes or parts thereof of various organisms. Determining the detailed physiological response properties of all the neurons in these connectomes is out of reach of current experimental technology. It is therefore unclear, to what extent knowledge of the connectome alone will advance a mechanistic understanding of computation occurring in these neuronal circuits, especially when the high-level function(s) of the said circuit is unknown.We are pursuing a research program to build theory in order to investigate these issues. In previously published work [1], towards this end, we have developed a theory of connectomic constraints for feedforward networks of neurons. Specifically, for feedforward networks equipped with neurons that obey a deterministic spiking neuron model, we asked if just by knowing the structure of a network, we could rule out spike-timed computations that it could be doing, no matter what response properties each of its neurons may have. Our neurons were abstract mathematical objects that satisfied a small number of axioms that correspond to certain broadly-obeyed properties of neurons.Here, we develop additional theoretical tools and notions to address these questions. The idea is to study the space of all possible spike-train to spike-train transformations. We are interested in asking how the subset of transformations spanned by networks of specific architectures can be related to hierarchical subsets of the space that are characterized by particular mathematical properties of transformations. In particular, given such a hierarchy of subsets, what is the "smallest" subset that contains the set of transformations spanned by networks of a specific class of architectures? Even if one cannot establish such a subset exactly, proving bounds on it (according to the hierarchy) might offer insight. After setting up the mathematical framework to make these notions precise, we construct explicit classes of hierarchies and prove a number of such lower bounds.

show abstract

“…Due to the infeasibility of manually performing reconstruction of thousands of neurons 4 , automation of this task has emerged as a major practical requirement for revealing neural circuits in VEM datasets 5 . Significant progress in automated neuron reconstruction has been achieved through the development of machine-learning (ML) based image segmentation methods optimized for tracing of neurites in VEM data [6][7][8][9][10] . ML methods require training data that has traditionally been created for each specific acquisition context, which differ by choice of species, brain area, imaging method, staining protocol, and other sample-specific details.…”

Section: Introductionmentioning

confidence: 99%

Segmentation-Enhanced CycleGAN

Jain

2019

Preprint

View full text Add to dashboard Cite

Algorithmic reconstruction of neurons from volume electron microscopy data traditionally requires training machine learning models on dataset-specific ground truth annotations that are expensive and tedious to acquire. We enhanced the training procedure of an unsupervised image-to-image translation method with additional components derived from an automated neuron segmentation approach. We show that this method, Segmentation-Enhanced CycleGAN (SECGAN), enables near perfect reconstruction accuracy on a benchmark connectomics segmentation dataset despite operating in a "zero-shot" setting in which the segmentation model was trained using only volumetric labels from a different dataset and imaging method. By reducing or eliminating the need for novel ground truth annotations, SECGANs alleviate one of the main practical burdens involved in pursuing automated reconstruction of volume electron microscopy data. Neural network architecturesWe used a 3d generalization of ResNet-18 24 as the architecture for all three discriminator networks, and a stack of 8 residual convolutional modules as the architecture of the generator networks. We also tested a U-net-style generator, as well as a simple convolution-pooling network as the discriminator, and found these alternatives resulted in worse segmentation quality.

show abstract

Multicut brings automated neurite segmentation closer to human performance

Cited by 145 publications

References 2 publications

EM connectomics reveals axonal target variation in a sequence-generating network

EM connectomics reveals axonal target variation in a sequence-generating network

Hierarchies & Lower Bounds in Theoretical Connectomics

Segmentation-Enhanced CycleGAN

Contact Info

Product

Resources

About