Category: Frontiers in Neuroinformatics

Advancements in multichannel recordings of single-unit activity (SUA) in vivo present an opportunity to discover novel features of spatially-varying extracellularly-recorded action potentials (EAPs) that are useful for identifying neuron-types. Traditional approaches to classifying neuron-types often rely on computing EAP waveform features based on conventions of single-channel recordings and thus inherit their limitations. However, spatiotemporal EAP waveforms are the product of signals from underlying current sources being mixed within the extracellular space. We introduce a machine learning approach to demix the underlying sources of spatiotemporal EAP waveforms. Using biophysically realistic computational models, we simulate EAP waveforms and characterize them by the relative prevalence of these sources, which we use as features for identifying the neuron-types corresponding to recorded single units. These EAP sources have distinct spatial and multi-resolution temporal patterns that are robust to various sampling biases. EAP sources also are shared across many neuron-types, are predictive of gross morphological features, and expose underlying morphological domains. We then organize known neuron-types into a hierarchy of latent morpho-electrophysiological types based on differences in the source prevalences, which provides a multi-level classification scheme. We validate the robustness, accuracy, and interpretations of our demixing approach by analyzing simulated EAPs from morphologically detailed models with classification and clustering methods. This simulation-based approach provides a machine learning strategy for neuron-type identification.

The study presents a novel approach designed to detect time-continuous states in time-series data, called the State-Detecting Algorithm (SDA). The SDA operates on unlabeled data and detects optimal change-points among intrinsic functional states in time-series data based on an ensemble of Ward's hierarchical clustering with time-connectivity constraint. The algorithm chooses the best number of states and optimal state boundaries, maximizing clustering quality metrics. We also introduce a series of methods to estimate the performance and confidence of the SDA when the ground truth annotation is unavailable. These include information value analysis, paired statistical tests, and predictive modeling analysis. The SDA was validated on EEG recordings of Guhyasamaja meditation practice with a strict staged protocol performed by three experienced Buddhist practitioners in an ecological setup. The SDA used neurophysiological descriptors as inputs, including PSD, power indices, coherence, and PLV. Post-hoc analysis of the obtained EEG states revealed significant differences compared to the baseline and neighboring states. The SDA was found to be stable with respect to state order organization and showed poor clustering quality metrics and no statistical significance between states when applied to randomly shuffled epochs (i.e., surrogate subject data used as controls). The SDA can be considered a general data-driven approach that detects hidden functional states associated with the mental processes evolving during meditation or other ongoing mental and cognitive processes.

Neuroscience has made significant strides over the past decade in moving from a largely closed science characterized by anemic data sharing, to a largely open science where the amount of publicly available neuroscience data has increased dramatically. While this increase is driven in significant part by large prospective data sharing studies, we are starting to see increased sharing in the long tail of neuroscience data, driven no doubt by journal requirements and funder mandates. Concomitant with this shift to open is the increasing support of the FAIR data principles by neuroscience practices and infrastructure. FAIR is particularly critical for neuroscience with its multiplicity of data types, scales and model systems and the infrastructure that serves them. As envisioned from the early days of neuroinformatics, neuroscience is currently served by a globally distributed ecosystem of neuroscience-centric data repositories, largely specialized around data types. To make neuroscience data findable, accessible, interoperable, and reusable requires the coordination across different stakeholders, including the researchers who produce the data, data repositories who make it available, the aggregators and indexers who field search engines across the data, and community organizations who help to coordinate efforts and develop the community standards critical to FAIR. The International Neuroinformatics Coordinating Facility has led efforts to move neuroscience toward FAIR, fielding several resources to help researchers and repositories achieve FAIR. In this perspective, I provide an overview of the components and practices required to achieve FAIR in neuroscience and provide thoughts on the past, present and future of FAIR infrastructure for neuroscience, from the laboratory to the search engine.

Neural decoding is a powerful method to analyze neural activity. However, the code needed to run a decoding analysis can be complex, which can present a barrier to using the method. In this paper we introduce a package that makes it easy to perform decoding analyses in the R programing language. We describe how the package is designed in a modular fashion which allows researchers to easily implement a range of different analyses. We also discuss how to format data to be able to use the package, and we give two examples of how to use the package to analyze real data. We believe that this package, combined with the rich data analysis ecosystem in R, will make it significantly easier for researchers to create reproducible decoding analyses, which should help increase the pace of neuroscience discoveries.

Goal-driven deep learning increasingly supplements classical modeling approaches in computational neuroscience. The strength of deep neural networks as models of the brain lies in their ability to autonomously learn the connectivity required to solve complex and ecologically valid tasks, obviating the need for hand-engineered or hypothesis-driven connectivity patterns. Consequently, goal-driven models can generate hypotheses about the neurocomputations underlying cortical processing that are grounded in macro- and mesoscopic anatomical properties of the network's biological counterpart. Whereas, goal-driven modeling is already becoming prevalent in the neuroscience of perception, its application to the sensorimotor domain is currently hampered by the complexity of the methods required to train models comprising the closed sensation-action loop. This paper describes AngoraPy, a Python library that mitigates this obstacle by providing researchers with the tools necessary to train complex recurrent convolutional neural networks that model the human sensorimotor system. To make the technical details of this toolkit more approachable, an illustrative example that trains a recurrent toy model on in-hand object manipulation accompanies the theoretical remarks. An extensive benchmark on various classical, 3D robotic, and anthropomorphic control tasks demonstrates AngoraPy's general applicability to a wide range of tasks. Together with its ability to adaptively handle custom architectures, the flexibility of this toolkit demonstrates its power for goal-driven sensorimotor modeling.

The main olfactory bulb is the key element of the olfactory pathway of rodents. To precisely dissect the neural pathway in the main olfactory bulb (MOB), it is necessary to construct the three-dimensional morphologies of the anatomical structures within it with micro-level resolution. However, the construction remains challenging due to the complicated shape of the anatomical structures in the main olfactory bulb and the high resolution of micro-optical images. To address these issues, we propose an interactive volume image segmentation method with micro-level resolution in the horizontal and axial direction. Firstly, we obtain the initial location of the anatomical structures by manual annotation and design a patch-based neural network to learn the complex texture feature of the anatomical structures. Then we randomly sample some patches to predict by the trained network and perform an annotation reconstruction based on intensity calculation to get the final location results of the anatomical structures. Our experiments were conducted using Nissl-stained brain images acquired by the Micro-optical sectioning tomography (MOST) system. Our method achieved a mean dice similarity coefficient (DSC) of 81.8% and obtain the best segmentation performance. At the same time, the experiment shows the three-dimensional morphology reconstruction results of the anatomical structures in the main olfactory bulb are smooth and consistent with their natural shapes, which addresses the possibility of constructing three-dimensional morphologies of the anatomical structures in the whole brain.