Neural networks, composed of many neurons and governed by complex interactions between them, are a widely accepted formalism for modeling and exploring global dynamics and emergent properties in brain systems. In the past decades, experimental evidence of computationally relevant neuron-astrocyte interactions, as well as the astrocytic modulation of global neural dynamics, have accumulated. These findings motivated advances in computational glioscience and inspired several models integrating mechanisms of neuron-astrocyte interactions into the standard neural network formalism. These models were developed to study, for example, synchronization, information transfer, synaptic plasticity, and hyperexcitability, as well as classification tasks and hardware implementations. We here focus on network models of at least two neurons interacting bidirectionally with at least two astrocytes that include explicitly modeled astrocytic calcium dynamics. In this study, we analyze the evolution of these models and the biophysical, biochemical, cellular, and network mechanisms used to construct them. Based on our analysis, we propose how to systematically describe and categorize interaction schemes between cells in neuron-astrocyte networks. We additionally study the models in view of the existing experimental data and present future perspectives. Our analysis is an important first step towards understanding astrocytic contribution to brain functions. However, more advances are needed to collect comprehensive data about astrocyte morphology and physiology in vivo and to better integrate them in data-driven computational models. Broadening the discussion about theoretical approaches and expanding the computational tools is necessary to better understand astrocytes' roles in brain functions.

Quantitative analysis of white matter fiber tracts from diffusion Magnetic Resonance Imaging (dMRI) data is of great significance in health and disease. For example, analysis of fiber tracts related to anatomically meaningful fiber bundles is highly demanded in pre-surgical and treatment planning, and the surgery outcome depends on accurate segmentation of the desired tracts. Currently, this process is mainly done through time-consuming manual identification performed by neuro-anatomical experts. However, there is a broad interest in automating the pipeline such that it is fast, accurate, and easy to apply in clinical settings and also eliminates the intra-reader variabilities. Following the advancements in medical image analysis using deep learning techniques, there has been a growing interest in using these techniques for the task of tract identification as well. Recent reports on this application show that deep learning-based tract identification approaches outperform existing state-of-the-art methods. This paper presents a review of current tract identification approaches based on deep neural networks. First, we review the recent deep learning methods for tract identification. Next, we compare them with respect to their performance, training process, and network properties. Finally, we end with a critical discussion of open challenges and possible directions for future works.

Deep learning algorithms have a huge influence on tackling research issues in the field of medical image processing. It acts as a vital aid for the radiologists in producing accurate results toward effective disease diagnosis. The objective of this research is to highlight the importance of deep learning models in the detection of Alzheimer's Disease (AD). The main objective of this research is to analyze different deep learning methods used for detecting AD. This study examines 103 research articles published in various research databases. These articles have been selected based on specific criteria to find the most relevant findings in the field of AD detection. The review was carried out based on deep learning techniques such as Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and Transfer Learning (TL). To propose accurate methods for the detection, segmentation, and severity grading of AD, the radiological features need to be examined in greater depth. This review attempts to analyze different deep learning methods applied for AD detection using neuroimaging modalities like Positron Emission Tomography (PET), Magnetic Resonance Imaging (MRI), etc. The focus of this review is restricted to deep learning works based on radiological imaging data for AD detection. There are a few works that have utilized other biomarkers to understand the effect of AD. Also, articles published in English were alone considered for analysis. This work concludes by highlighting the key research issues towards effective AD detection. Though several methods have yielded promising results in AD detection, the progression from Mild Cognitive Impairment (MCI) to AD need to be analyzed in greater depth using DL models.

Magnetic resonance imaging (MRI) and light-sheet fluorescence microscopy (LSFM) are technologies that enable non-disruptive 3-dimensional imaging of whole mouse brains. A combination of complementary information from both modalities is desirable for studying neuroscience in general, disease progression and drug efficacy. Although both technologies rely on atlas mapping for quantitative analyses, the translation of LSFM recorded data to MRI templates has been complicated by the morphological changes inflicted by tissue clearing and the enormous size of the raw data sets. Consequently, there is an unmet need for tools that will facilitate fast and accurate translation of LSFM recorded brains to in vivo, non-distorted templates. In this study, we have developed a bidirectional multimodal atlas framework that includes brain templates based on both imaging modalities, region delineations from the Allen's Common Coordinate Framework, and a skull-derived stereotaxic coordinate system. The framework also provides algorithms for bidirectional transformation of results obtained using either MR or LSFM (iDISCO cleared) mouse brain imaging while the coordinate system enables users to easily assign in vivo coordinates across the different brain templates.

Head CT, which includes the facial region, can visualize faces using 3D reconstruction, raising concern that individuals may be identified. We developed a new de-identification technique that distorts the faces of head CT images. Head CT images that were distorted were labeled as "original images" and the others as "reference images." Reconstructed face models of both were created, with 400 control points on the facial surfaces. All voxel positions in the original image were moved and deformed according to the deformation vectors required to move to corresponding control points on the reference image. Three face detection and identification programs were used to determine face detection rates and match confidence scores. Intracranial volume equivalence tests were performed before and after deformation, and correlation coefficients between intracranial pixel value histograms were calculated. Output accuracy of the deep learning model for intracranial segmentation was determined using Dice Similarity Coefficient before and after deformation. The face detection rate was 100%, and match confidence scores were < 90. Equivalence testing of the intracranial volume revealed statistical equivalence before and after deformation. The median correlation coefficient between intracranial pixel value histograms before and after deformation was 0.9965, indicating high similarity. Dice Similarity Coefficient values of original and deformed images were statistically equivalent. We developed a technique to de-identify head CT images while maintaining the accuracy of deep-learning models. The technique involves deforming images to prevent face identification, with minimal changes to the original information.

Activation likelihood estimation (ALE) is among the most used algorithms to perform neuroimaging meta-analysis. Since its first implementation, several thresholding procedures had been proposed, all referred to the frequentist framework, returning a rejection criterion for the null hypothesis according to the critical p-value selected. However, this is not informative in terms of probabilities of the validity of the hypotheses. Here, we describe an innovative thresholding procedure based on the concept of minimum Bayes factor (mBF). The use of the Bayesian framework allows to consider different levels of probability, each of these being equally significant. In order to simplify the translation between the common ALE practice and the proposed approach, we analised six task-fMRI/VBM datasets and determined the mBF values equivalent to the currently recommended frequentist thresholds based on Family Wise Error (FWE). Sensitivity and robustness toward spurious findings were also analyzed. Results showed that the cutoff log10(mBF) = 5 is equivalent to the FWE threshold, often referred as voxel-level threshold, while the cutoff log10(mBF) = 2 is equivalent to the cluster-level FWE (c-FWE) threshold. However, only in the latter case voxels spatially far from the blobs of effect in the c-FWE ALE map survived. Therefore, when using the Bayesian thresholding the cutoff log10(mBF) = 5 should be preferred. However, being in the Bayesian framework, lower values are all equally significant, while suggesting weaker level of force for that hypothesis. Hence, results obtained through less conservative thresholds can be legitimately discussed without losing statistical rigor. The proposed technique adds therefore a powerful tool to the human-brain-mapping field.

Fetal functional magnetic resonance imaging (fMRI) offers critical insight into the developing brain and could aid in predicting developmental outcomes. As the fetal brain is surrounded by heterogeneous tissue, it is not possible to use adult- or child-based segmentation toolboxes. Manually-segmented masks can be used to extract the fetal brain; however, this comes at significant time costs. Here, we present a new BIDS App for masking fetal fMRI, funcmasker-flex, that overcomes these issues with a robust 3D convolutional neural network (U-net) architecture implemented in an extensible and transparent Snakemake workflow. Open-access fetal fMRI data with manual brain masks from 159 fetuses (1103 total volumes) were used for training and testing the U-net model. We also tested generalizability of the model using 82 locally acquired functional scans from 19 fetuses, which included over 2300 manually segmented volumes. Dice metrics were used to compare performance of funcmasker-flex to the ground truth manually segmented volumes, and segmentations were consistently robust (all Dice metrics ≥ 0.74). The tool is freely available and can be applied to any BIDS dataset containing fetal bold sequences. Funcmasker-flex reduces the need for manual segmentation, even when applied to novel fetal functional datasets, resulting in significant time-cost savings for performing fetal fMRI analysis.

To accurately explore the anatomical organization of neural circuits in the brain, it is crucial to map the experimental brain data onto a standardized system of coordinates. Studying 2D histological mouse brain slices remains the standard procedure in many laboratories. Mapping these 2D brain slices is challenging; due to deformations, artifacts, and tilted angles introduced during the standard preparation and slicing process. In addition, analysis of experimental mouse brain slices can be highly dependent on the level of expertise of the human operator. Here we propose a computational tool for Accurate Mouse Brain Image Analysis (AMBIA), to map 2D mouse brain slices on the 3D brain model with minimal human intervention. AMBIA has a modular design that comprises a localization module and a registration module. The localization module is a deep learning-based pipeline that localizes a single 2D slice in the 3D Allen Brain Atlas and generates a corresponding atlas plane. The registration module is built upon the Ardent python package that performs deformable 2D registration between the brain slice to its corresponding atlas. By comparing AMBIA's performance in localization and registration to human ratings, we demonstrate that it performs at a human expert level. AMBIA provides an intuitive and highly efficient way for accurate registration of experimental 2D mouse brain images to 3D digital mouse brain atlas. Our tool provides a graphical user interface and it is designed to be used by researchers with minimal programming knowledge.

The sharing of open-access neuroimaging data has increased significantly during the last few years. Sharing neuroimaging data is crucial to accelerating scientific advancement, particularly in the field of neuroscience. A number of big initiatives that will increase the amount of available neuroimaging data are currently in development. The Big Brain Data Initiative project was started by Universiti Sains Malaysia as the first neuroimaging data repository platform in Malaysia for the purpose of data sharing. In order to ensure that the neuroimaging data in this project is accessible, usable, and secure, as well as to offer users high-quality data that can be consistently accessed, we first came up with good data stewardship practices. Then, we developed MyneuroDB, an online repository database system for data sharing purposes. Here, we describe the Big Brain Data Initiative and MyneuroDB, a data repository that provides the ability to openly share neuroimaging data, currently including magnetic resonance imaging (MRI), electroencephalography (EEG), and magnetoencephalography (MEG), following the FAIR principles for data sharing.

Traumatic brain injury (TBI) and repetitive head impacts can result in a wide range of neurological symptoms. Despite being the most common neurological disorder in the world, repeat head impacts and TBI do not have any FDA-approved treatments. Single neuron modeling allows researchers to extrapolate cellular changes in individual neurons based on experimental data. We recently characterized a model of high frequency head impact (HFHI) with a phenotype of cognitive deficits associated with decreases in neuronal excitability of CA1 neurons and synaptic changes. While the synaptic changes have been interrogated in vivo, the cause and potential therapeutic targets of hypoexcitability following repetitive head impacts are unknown. Here, we generated in silico models of CA1 pyramidal neurons from current clamp data of control mice and mice that sustained HFHI. We use a directed evolution algorithm with a crowding penalty to generate a large and unbiased population of plausible models for each group that approximated the experimental features. The HFHI neuron model population showed decreased voltage gated sodium conductance and a general increase in potassium channel conductance. We used partial least squares regression analysis to identify combinations of channels that may account for CA1 hypoexcitability after HFHI. The hypoexcitability phenotype in models was linked to A- and M-type potassium channels in combination, but not by any single channel correlations. We provide an open access set of CA1 pyramidal neuron models for both control and HFHI conditions that can be used to predict the effects of pharmacological interventions in TBI models.

In this paper we demonstrate a generalized and simplified pipeline called axonal spectrum imaging (AxSI) for in-vivo estimation of axonal characteristics in the human brain. Whole-brain estimation of the axon diameter, in-vivo and non-invasively, across all fiber systems will allow exploring uncharted aspects of brain structure and function relations with emphasis on connectivity and connectome analysis. While axon diameter mapping is important in and of itself, its correlation with conduction velocity will allow, for the first time, the explorations of information transfer mechanisms within the brain. We demonstrate various well-known aspects of axonal morphometry (e.g., the corpus callosum axon diameter variation) as well as other aspects that are less explored (e.g., axon diameter-based separation of the superior longitudinal fasciculus into segments). Moreover, we have created an MNI based mean axon diameter map over the entire brain for a large cohort of subjects providing the reference basis for future studies exploring relation between axon properties, its connectome representation, and other functional and behavioral aspects of the brain.

Understanding functions of astrocytes can be greatly enhanced by building and simulating computational models that capture their morphological details. Novel computational tools enable utilization of existing morphological data of astrocytes and building models that have appropriate level of details for specific simulation purposes. In addition to analyzing existing computational tools for constructing, transforming, and assessing astrocyte morphologies, we present here the CellRemorph toolkit implemented as an add-on for Blender, a 3D modeling platform increasingly recognized for its utility for manipulating 3D biological data. To our knowledge, CellRemorph is the first toolkit for transforming astrocyte morphologies from polygonal surface meshes into adjustable surface point clouds and vice versa, precisely selecting nanoprocesses, and slicing morphologies into segments with equal surface areas or volumes. CellRemorph is an open-source toolkit under the GNU General Public License and easily accessible via an intuitive graphical user interface. CellRemorph will be a valuable addition to other Blender add-ons, providing novel functionality that facilitates the creation of realistic astrocyte morphologies for different types of morphologically detailed simulations elucidating the role of astrocytes both in health and disease.

Data from multisite magnetic resonance imaging (MRI) studies contain variance attributable to the scanner that can reduce statistical power and potentially bias results if not appropriately managed. The Adolescent Cognitive Brain Development (ABCD) study is an ongoing, longitudinal neuroimaging study acquiring data from over 11,000 children starting at 9-10 years of age. These scans are acquired on 29 different scanners of 5 different model types manufactured by 3 different vendors. Publicly available data from the ABCD study include structural MRI (sMRI) measures such as cortical thickness and diffusion MRI (dMRI) measures such as fractional anisotropy. In this work, we 1) quantify the variance attributable to scanner effects in the sMRI and dMRI datasets, 2) demonstrate the effectiveness of the data harmonization approach called ComBat to address scanner effects, and 3) present a simple, open-source tool for investigators to harmonize image features from the ABCD study. Scanner-induced variance was present in every image feature and varied in magnitude by feature type and brain location. For almost all features, scanner variance exceeded variability attributable to age and sex. ComBat harmonization was shown to effectively remove scanner induced variance from all image features while preserving the biological variability in the data. Moreover, we show that for studies examining relatively small subsamples of the ABCD dataset, the use of ComBat harmonized data provides more accurate estimates of effect sizes compared to controlling for scanner effects using ordinary least squares regression.