ManiVault: A Flexible and Extensible Visual Analytics Framework for High-Dimensional Data.

Exploration and analysis of high-dimensional data are important tasks in many fields that produce large and complex data, like the financial sector, systems biology, or cultural heritage. Tailor-made visual analytics software is developed for each specific application, limiting their applicability in other fields. However, as diverse as these fields are, their characteristics and requirements for data analysis are conceptually similar. Many applications share abstract tasks and data types and are often constructed with similar building blocks. Developing such applications, even when based mostly on existing building blocks, requires significant engineering efforts. We developed ManiVault, a flexible and extensible open-source visual analytics framework for analyzing high-dimensional data. The primary objective of ManiVault is to facilitate rapid prototyping of visual analytics workflows for visualization software developers and practitioners alike. ManiVault is built using a plugin-based architecture that offers easy extensibility. While our architecture deliberately keeps plugins self-contained, to guarantee maximum flexibility and re-usability, we have designed and implemented a messaging API for tight integration and linking of modules to support common visual analytics design patterns. We provide several visualization and analytics plugins, and ManiVault's API makes the integration of new plugins easy for developers. ManiVault facilitates the distribution of visualization and analysis pipelines and results for practitioners through saving and reproducing complete application states. As such, ManiVault can be used as a communication tool among researchers to discuss workflows and results. A copy of this paper and all supplemental material is available at osf.io/9k6jw, and source code at github.com/ManiVaultStudio.

Comparative cellular analysis of motor cortex in human, marmoset and mouse.

The primary motor cortex (M1) is essential for voluntary fine-motor control and is functionally conserved across mammals1. Here, using high-throughput transcriptomic and epigenomic profiling of more than 450,000 single nuclei in humans, marmoset monkeys and mice, we demonstrate a broadly conserved cellular makeup of this region, with similarities that mirror evolutionary distance and are consistent between the transcriptome and epigenome. The core conserved molecular identities of neuronal and non-neuronal cell types allow us to generate a cross-species consensus classification of cell types, and to infer conserved properties of cell types across species. Despite the overall conservation, however, many species-dependent specializations are apparent, including differences in cell-type proportions, gene expression, DNA methylation and chromatin state. Few cell-type marker genes are conserved across species, revealing a short list of candidate genes and regulatory mechanisms that are responsible for conserved features of homologous cell types, such as the GABAergic chandelier cells. This consensus transcriptomic classification allows us to use patch-seq (a combination of whole-cell patch-clamp recordings, RNA sequencing and morphological characterization) to identify corticospinal Betz cells from layer 5 in non-human primates and humans, and to characterize their highly specialized physiology and anatomy. These findings highlight the robust molecular underpinnings of cell-type diversity in M1 across mammals, and point to the genes and regulatory pathways responsible for the functional identity of cell types and their species-specific adaptations.

GPGPU Linear Complexity t-SNE Optimization.

In recent years the t-distributed Stochastic Neighbor Embedding (t-SNE) algorithm has become one of the most used and insightful techniques for exploratory data analysis of high-dimensional data. It reveals clusters of high-dimensional data points at different scales while only requiring minimal tuning of its parameters. However, the computational complexity of the algorithm limits its application to relatively small datasets. To address this problem, several evolutions of t-SNE have been developed in recent years, mainly focusing on the scalability of the similarity computations between data points. However, these contributions are insufficient to achieve interactive rates when visualizing the evolution of the t-SNE embedding for large datasets. In this work, we present a novel approach to the minimization of the t-SNE objective function that heavily relies on graphics hardware and has linear computational complexity. Our technique decreases the computational cost of running t-SNE on datasets by orders of magnitude and retains or improves on the accuracy of past approximated techniques. We propose to approximate the repulsive forces between data points by splatting kernel textures for each data point. This approximation allows us to reformulate the t-SNE minimization problem as a series of tensor operations that can be efficiently executed on the graphics card. An efficient implementation of our technique is integrated and available for use in the widely used Google TensorFlow.js, and an open-source C++ library.

Microstructural white matter changes preceding white matter hyperintensities in migraine.

OBJECTIVE: We used magnetization transfer imaging to assess white matter tissue integrity in migraine, to explore whether white matter microstructure was more diffusely affected beyond visible white matter hyperintensities (WMHs), and to explore whether focal invisible microstructural changes precede visible focal WMHs in migraineurs. METHODS: We included 137 migraineurs (79 with aura, 58 without aura) and 74 controls from the Cerebral Abnormalities in Migraine, an Epidemiological Risk Analysis (CAMERA) study, a longitudinal population-based study on structural brain lesions in migraine patients, who were scanned at baseline and at a 9-year follow-up. To assess microstructural brain tissue integrity, baseline magnetization transfer ratio (MTR) values were calculated for whole brain white matter. Baseline MTR values were determined for areas of normal-appearing white matter (NAWM) that had progressed into MRI-detectable WMHs at follow-up and compared to MTR values of contralateral NAWM. RESULTS: MTR values for whole brain white matter did not differ between migraineurs and controls. In migraineurs, but not in controls, NAWM that later progressed to WMHs at follow-up had lower mean MTR (mean [SD] 0.354 [0.009] vs 0.356 [0.008], p = 0.047) at baseline as compared to contralateral white matter. CONCLUSIONS: We did not find evidence for widespread microstructural white matter changes in migraineurs compared to controls. However, our findings suggest that a gradual or stepwise process might be responsible for evolution of focal invisible microstructural changes into focal migraine-related visible WMHs.

In vivo assessment of iron content of the cerebral cortex in healthy aging using 7-Tesla T2*-weighted phase imaging.

Accumulation of brain iron has been suggested as a biomarker of neurodegeneration. Increased iron has been seen in the cerebral cortex in postmortem studies of neurodegenerative diseases and healthy aging. Until recently, the diminutive thickness of the cortex and its relatively low iron content have hampered in vivo study of cortical iron accumulation. Using phase images of a T2*-weighted sequence at ultrahigh field strength (7 Tesla), we examined the iron content of 22 cortical regions in 70 healthy subjects aged 22-80 years. The cortex was automatically segmented and parcellated, and phase shift was analyzed using an in-house developed method. We found a significant increase in phase shift with age in 20 of 22 cortical regions, concurrent with current understanding of cortical iron accumulation. Our findings suggest that increased cortical iron content can be assessed in healthy aging in vivo. The high spatial resolution and sensitivity to iron of our method make it a potentially useful tool for studying cortical iron accumulation in healthy aging and neurodegenerative diseases.

BrainScope: interactive visual exploration of the spatial and temporal human brain transcriptome.

Spatial and temporal brain transcriptomics has recently emerged as an invaluable data source for molecular neuroscience. The complexity of such data poses considerable challenges for analysis and visualization. We present BrainScope: a web portal for fast, interactive visual exploration of the Allen Atlases of the adult and developing human brain transcriptome. Through a novel methodology to explore high-dimensional data (dual t-SNE), BrainScope enables the linked, all-in-one visualization of genes and samples across the whole brain and genome, and across developmental stages. We show that densities in t-SNE scatter plots of the spatial samples coincide with anatomical regions, and that densities in t-SNE scatter plots of the genes represent gene co-expression modules that are significantly enriched for biological functions. We also show that the topography of the gene t-SNE maps reflect brain region-specific gene functions, enabling hypothesis and data driven research. We demonstrate the discovery potential of BrainScope through three examples: (i) analysis of cell type specific gene sets, (ii) analysis of a set of stable gene co-expression modules across the adult human donors and (iii) analysis of the evolution of co-expression of oligodendrocyte specific genes over developmental stages. BrainScope is publicly accessible at www.brainscope.nl.

Fast Automatic Step Size Estimation for Gradient Descent Optimization of Image Registration.

Fast automatic image registration is an important prerequisite for image-guided clinical procedures. However, due to the large number of voxels in an image and the complexity of registration algorithms, this process is often very slow. Stochastic gradient descent is a powerful method to iteratively solve the registration problem, but relies for convergence on a proper selection of the optimization step size. This selection is difficult to perform manually, since it depends on the input data, similarity measure and transformation model. The Adaptive Stochastic Gradient Descent (ASGD) method is an automatic approach, but it comes at a high computational cost. In this paper, we propose a new computationally efficient method (fast ASGD) to automatically determine the step size for gradient descent methods, by considering the observed distribution of the voxel displacements between iterations. A relation between the step size and the expectation and variance of the observed distribution is derived. While ASGD has quadratic complexity with respect to the transformation parameters, fast ASGD only has linear complexity. Extensive validation has been performed on different datasets with different modalities, inter/intra subjects, different similarity measures and transformation models. For all experiments, we obtained similar accuracy as ASGD. Moreover, the estimation time of fast ASGD is reduced to a very small value, from 40 s to less than 1 s when the number of parameters is 105, almost 40 times faster. Depending on the registration settings, the total registration time is reduced by a factor of 2.5-7 × for the experiments in this paper.

Hippocampal atrophy in people with memory deficits: results from the population-based IPREA study.

BACKGROUND: Clinical studies have shown that hippocampal atrophy is present before dementia in people with memory deficits and can predict dementia development. The question remains whether this association holds in the general population. This is of interest for the possible use of hippocampal atrophy to screen population for preventive interventions. The aim of this study was to assess hippocampal volume and shape abnormalities in elderly adults with memory deficits in a cross-sectional population-based study. METHODS: We included individuals participating in the Italian Project on the Epidemiology of Alzheimer Disease (IPREA) study: 75 cognitively normal individuals (HC), 31 individuals with memory deficits (MEM), and 31 individuals with memory deficits not otherwise specified (MEMnos). Hippocampal volumes and shape were extracted through manual tracing and the growing and adaptive meshes (GAMEs) shape-modeling algorithm. We investigated between-group differences in hippocampal volume and shape, and correlations with memory deficits. RESULTS: In MEM participants, hippocampal volumes were significantly smaller than in HC and were mildly associated with worse memory scores. Memory-associated shape changes mapped to the anterior hippocampus. Shape-based analysis detected no significant difference between MEM and HC, while MEMnos showed shape changes in the posterior hippocampus compared with HC and MEM groups. CONCLUSIONS: These findings support the discriminant validity of hippocampal volumetry as a biomarker of memory impairment in the general population. The detection of shape changes in MEMnos but not in MEM participants suggests that shape-based biomarkers might lack sensitivity to detect Alzheimer's-like pathology in the general population.

Structural brain changes in migraine.

CONTEXT: A previous cross-sectional study showed an association of migraine with a higher prevalence of magnetic resonance imaging (MRI)-measured ischemic lesions in the brain. OBJECTIVE: To determine whether women or men with migraine (with and without aura) have a higher incidence of brain lesions 9 years after initial MRI, whether migraine frequency was associated with progression of brain lesions, and whether progression of brain lesions was associated with cognitive decline. DESIGN, SETTING, AND PARTICIPANTS: In a follow-up of the 2000 Cerebral Abnormalities in Migraine, an Epidemiological Risk Analysis cohort, a prospective population-based observational study of Dutch participants with migraine and an age- and sex-matched control group, 203 of the 295 baseline participants in the migraine group and 83 of 140 in the control group underwent MRI scan in 2009 to identify progression of MRI-measured brain lesions. Comparisons were adjusted for age, sex, hypertension, diabetes, and educational level. The participants in the migraine group were a mean 57 years (range, 43-72 years), and 71% were women. Those in the control group were a mean 55 years (range, 44-71 years), and 69% were women. MAIN OUTCOME MEASURES Progression of MRI-measured cerebral deep white matter hyperintensities, infratentorial hyperintensities, and posterior circulation territory infarctlike lesions. Change in cognition was also measured. RESULTS: Of the 145 women in the migraine group, 112 (77%) vs 33 of 55 women (60%) in the control group had progression of deep white matter hyperintensities (adjusted odds ratio [OR], 2.1; 95%CI, 1.0-4.1; P = .04). There were no significant associations of migraine with progression of infratentorial hyperintensities: 21 participants (15%) in the migraine group and 1 of 57 participants (2%) in the control group showed progression (adjusted OR, 7.7; 95% CI, 1.0-59.5; P = .05) or new posterior circulation territory infarctlike lesions: 10 of 203 participants (5%) in the migraine group but none of 83 in the control group (P = .07). There was no association of number or frequency of migraine headaches with progression of lesions. There was no significant association of high vs nonhigh deep white matter hyperintensity load with change in cognitive scores (-3.7 in the migraine group vs 1.4 in the control group; 95% CI, -4.4 to 0.2; adjusted P = .07). CONCLUSIONS: In a community-based cohort followed up after 9 years, women with migraine had a higher incidence of deep white matter hyperintensities but did not have significantly higher progression of other MRI-measured brain changes. There was no association of migraine with progression of any MRI-measured brain lesions in men.

Non-parametric model selection for subject-specific topological organization of resting-state functional connectivity.

In recent years, graph theory has been successfully applied to study functional and anatomical connectivity networks in the human brain. Most of these networks have shown small-world topological characteristics: high efficiency in long distance communication between nodes, combined with highly interconnected local clusters of nodes. Moreover, functional studies performed at high resolutions have presented convincing evidence that resting-state functional connectivity networks exhibits (exponentially truncated) scale-free behavior. Such evidence, however, was mostly presented qualitatively, in terms of linear regressions of the degree distributions on log-log plots. Even when quantitative measures were given, these were usually limited to the r(2) correlation coefficient. However, the r(2) statistic is not an optimal estimator of explained variance, when dealing with (truncated) power-law models. Recent developments in statistics have introduced new non-parametric approaches, based on the Kolmogorov-Smirnov test, for the problem of model selection. In this work, we have built on this idea to statistically tackle the issue of model selection for the degree distribution of functional connectivity at rest. The analysis, performed at voxel level and in a subject-specific fashion, confirmed the superiority of a truncated power-law model, showing high consistency across subjects. Moreover, the most highly connected voxels were found to be consistently part of the default mode network. Our results provide statistically sound support to the evidence previously presented in literature for a truncated power-law model of resting-state functional connectivity.