Moritz Helmstaedter

Petilla terminology: nomenclature of features of GABAergic interneurons of the cerebral cortex.

Neuroscience produces a vast amount of data from an enormous diversity of neurons. A neuronal classification system is essential to organize such data and the knowledge that is derived from them. Classification depends on the unequivocal identification of the features that distinguish one type of neuron from another. The problems inherent in this are particularly acute when studying cortical interneurons. To tackle this, we convened a representative group of researchers to agree on a set of terms to describe the anatomical, physiological and molecular features of GABAergic interneurons of the cerebral cortex. The resulting terminology might provide a stepping stone towards a future classification of these complex and heterogeneous cells. Consistent adoption will be important for the success of such an initiative, and we also encourage the active involvement of the broader scientific community in the dynamic evolution of this project.

Wiring specificity in the direction-selectivity circuit of the retina

The proper connectivity between neurons is essential for the implementation of the algorithms used in neural computations, such as the detection of directed motion by the retina. The analysis of neuronal connectivity is possible with electron microscopy, but technological limitations have impeded the acquisition of high-resolution data on a large enough scale. Here we show, using serial block-face electron microscopy and two-photon calcium imaging, that the dendrites of mouse starburst amacrine cells make highly specific synapses with direction-selective ganglion cells depending on the ganglion cell’s preferred direction. Our findings indicate that a structural (wiring) asymmetry contributes to the computation of direction selectivity. The nature of this asymmetry supports some models of direction selectivity and rules out others. It also puts constraints on the developmental mechanisms behind the formation of synaptic connections. Our study demonstrates how otherwise intractable neurobiological questions can be addressed by combining functional imaging with the analysis of neuronal connectivity using large-scale electron microscopy.

Connectomic reconstruction of the inner plexiform layer in the mouse retina

Comprehensive high-resolution structural maps are central to functional exploration and understanding in biology. For the nervous system, in which high resolution and large spatial extent are both needed, such maps are scarce as they challenge data acquisition and analysis capabilities. Here we present for the mouse inner plexiform layer—the main computational neuropil region in the mammalian retina—the dense reconstruction of 950 neurons and their mutual contacts. This was achieved by applying a combination of crowd-sourced manual annotation and machine-learning-based volume segmentation to serial block-face electron microscopy data. We characterize a new type of retinal bipolar interneuron and show that we can subdivide a known type based on connectivity. Circuit motifs that emerge from our data indicate a functional mechanism for a known cellular response in a ganglion cell that detects localized motion, and predict that another ganglion cell is motion sensitive.

Convolutional networks can learn to generate affinity graphs for image segmentation

Many image segmentation algorithms first generate an affinity graph and then partition it. We present a machine learning approach to computing an affinity graph using a convolutional network (CN) trained using ground truth provided by human experts. The CN affinity graph can be paired with any standard partitioning algorithm and improves segmentation accuracy significantly compared to standard hand-designed affinity functions. We apply our algorithm to the challenging 3D segmentation problem of reconstructing neuronal processes from volumetric electron microscopy (EM) and show that we are able to learn a good affinity graph directly from the raw EM images. Further, we show that our affinity graph improves the segmentation accuracy of both simple and sophisticated graph partitioning algorithms. In contrast to previous work, we do not rely on prior knowledge in the form of hand-designed image features or image preprocessing. Thus, we expect our algorithm to generalize effectively to arbitrary image types.

Cell Type–Specific Three-Dimensional Structure of Thalamocortical Circuits in a Column of Rat Vibrissal Cortex

Soma location, dendrite morphology, and synaptic innervation may represent key determinants of functional responses of individual neurons, such as sensory-evoked spiking. Here, we reconstruct the 3D circuits formed by thalamocortical afferents from the lemniscal pathway and excitatory neurons of an anatomically defined cortical column in rat vibrissal cortex. We objectively classify 9 cortical cell types and estimate the number and distribution of their somata, dendrites, and thalamocortical synapses. Somata and dendrites of most cell types intermingle, while thalamocortical connectivity depends strongly upon the cell type and the 3D soma location of the postsynaptic neuron. Correlating dendrite morphology and thalamocortical connectivity to functional responses revealed that the lemniscal afferents can account for some of the cell type- and location-specific subthreshold and spiking responses after passive whisker touch (e.g., in layer 4, but not for other cell types, e.g., in layer 5). Our data provides a quantitative 3D prediction of the cell type–specific lemniscal synaptic wiring diagram and elucidates structure–function relationships of this physiologically relevant pathway at single-cell resolution.

High-accuracy neurite reconstruction for high-throughput neuroanatomy

Neuroanatomic analysis depends on the reconstruction of complete cell shapes. High-throughput reconstruction of neural circuits, or connectomics, using volume electron microscopy requires dense staining of all cells, which leads even experts to make annotation errors. Currently, reconstruction speed rather than acquisition speed limits the determination of neural wiring diagrams. We developed a method for fast and reliable reconstruction of densely labeled data sets. Our approach, based on manually skeletonizing each neurite redundantly (multiple times) with a visualization-annotation software tool called KNOSSOS, is ∼50-fold faster than volume labeling. Errors are detected and eliminated by a redundant-skeleton consensus procedure (RESCOP), which uses a statistical model of how true neurite connectivity is transformed into annotation decisions. RESCOP also estimates the reliability of consensus skeletons. Focused reannotation of difficult locations promises a rather steep increase of reliability as a function of the average skeleton redundancy and thus the nearly error-free analysis of large neuroanatomical datasets.

Supervised Learning of Image Restoration with Convolutional Networks

Convolutional networks have achieved a great deal of success in high-level vision problems such as object recognition. Here we show that they can also be used as a general method for low-level image processing. As an example of our approach, convolutional networks are trained using gradient learning to solve the problem of restoring noisy or degraded images. For our training data, we have used electron microscopic images of neural circuitry with ground truth restorations provided by human experts. On this dataset, Markov random field (MRF), conditional random field (CRF), and anisotropic diffusion algorithms perform about the same as simple thresholding, but superior performance is obtained with a convolutional network containing over 34,000 adjustable parameters. When restored by this convolutional network, the images are clean enough to be used for segmentation, whereas the other approaches fail in this respect. We do not believe that convolutional networks are fundamentally superior to MRFs as a representation for image processing algorithms. On the contrary, the two approaches are closely related. But in practice, it is possible to train complex convolutional networks, while even simple MRF models are hindered by problems with Bayesian learning and inference procedures. Our results suggest that high model complexity is the single most important factor for good performance, and this is possible with convolutional networks.

Cellular-resolution connectomics: challenges of dense neural circuit reconstruction

Neuronal networks are high-dimensional graphs that are packed into three-dimensional nervous tissue at extremely high density. Comprehensively mapping these networks is therefore a major challenge. Although recent developments in volume electron microscopy imaging have made data acquisition feasible for circuits comprising a few hundreds to a few thousands of neurons, data analysis is massively lagging behind. The aim of this perspective is to summarize and quantify the challenges for data analysis in cellular-resolution connectomics and describe current solutions involving online crowd-sourcing and machine-learning approaches.

3D structural imaging of the brain with photons and electrons.

Recent technological developments have renewed the interest in large-scale neural circuit reconstruction. To resolve the structure of entire circuits, thousands of neurons must be reconstructed and their synapses identified. Reconstruction techniques at the light microscopic level are capable of following sparsely labeled neurites over long distances, but fail with densely labeled neuropil. Electron microscopy provides the resolution required to resolve densely stained neuropil, but is challenged when data for volumes large enough to contain complete circuits need to be collected. Both photon-based and electron-based imaging methods will ultimately need highly automated data analysis, because the manual tracing of most networks of interest would require hundreds to tens of thousands of years in human labor.

Structural neurobiology: missing link to a mechanistic understanding of neural computation

High-resolution, comprehensive structural information is often the final arbiter between competing mechanistic models of biological processes, and can serve as inspiration for new hypotheses. In molecular biology, definitive structural data at atomic resolution are available for many macromolecules; however, information about the structure of the brain is much less complete, both in scope and resolution. Several technical developments over the past decade, such as serial block-face electron microscopy and trans-synaptic viral tracing, have made the structural biology of neural circuits conceivable: we may be able to obtain the structural information needed to reconstruct the network of cellular connections for large parts of, or even an entire, mouse brain within a decade or so. Given that the brains algorithms are ultimately encoded by this network, knowing where all of these connections are should, at the very least, provide the data needed to distinguish between models of neural computation.