Search

Journal of neural engineering 16(6), 066031 (2019). doi:10.1088/1741-2552/ab4113, Published by Institute of Physics Publishing, Bristol

SummarySparse population activity is a hallmark of supra-granular sensory neurons in neocortex. The mechanisms underlying sparseness are not well understood because a direct link between the neurons activated in vivo and their cellular properties investigated in vitro has been missing. We used two-photon calcium imaging to identify a subset of neurons in layer L2/3 (L2/3) of mouse primary somatosensory cortex that are highly active following principal whisker vibrotactile stimulation. These high responders were then tagged using photoconvertible green fluorescent protein for subsequent targeting in the brain slice using intracellular patch-clamp recordings and biocytin staining. This approach allowed us to investigate the structural and functional properties of high responders that distinguish them from less active control cells. Compared to less responsive L2/3 neurons, high responders displayed increased levels of stimulus-evoked and spontaneous activity, elevated noise and spontaneous pair-wise correlations, and stronger coupling to the population response. Intrinsic excitability was reduced in high responders, while other electrophysiological and morphological parameters were unchanged. Thus, the choice of which neurons participate in stimulus encoding may largely be determined by network connectivity rather than by cellular structure and function.

The rodent whisker system is a preferred model for studying plasticity in the somatosensory cortex (barrel cortex). Contrarily, only a small amount of research has been conducted to characterize the stability of neuronal population activity in the barrel cortex. We used the mouse whisker system to address the neuronal basis of stable perception in the somatosensory cortex. Cortical representation of periodic whisker deflections was studied in populations of neurons in supragranular layers over extended time periods (up to 3 months) with long-term two-photon Ca2+ imaging in anesthetized mice. We found that inmost of the neurons (87%), Ca2+ responses increased sublinearly with increasing number of contralateral whisker deflections. The imaged population of neurons was activated in a stereotypic way over days and for different deflection rates (pulse frequencies). Thus, pulse frequencies are coded by response strength rather than by distinct neuronal sub-populations. A small population of highly responsive neurons (similar to 3%) was sufficient to decode the whisker stimulus. This conserved functional map, led by a small set of highly responsive neurons, might form the foundation of stable sensory percepts. (C) 2015 Elsevier Inc. All rights reserved.

Stimulus-specific adaptation (SSA) to repetitive stimulation has been proposed to separate behaviorally relevant features from a stream of continuous sensory information. However, the exact mechanisms giving rise to SSA and cortical deviance detection are not well understood. We therefore used an oddball paradigm and multicontact electrodes to characterize single-neuron and local field potential responses to various deviant stimuli across the rat somatosensory cortex. Changing different single-whisker stimulus features evoked robust SSA in individual cortical neurons over a wide range of stimulus repetition rates (0.25–80 Hz). Notably, SSA was weakest in the granular input layer and significantly stronger in the supra- and infragranular layers, suggesting that a major part of SSA is generated within cortex. Moreover, we found a small subset of neurons in the granular layer with a deviant-specific late response, occurring roughly 200 ms after stimulus offset. This late deviant response exhibited true-deviance detection properties that were not explained by depression of sensory inputs. Our results show that deviant responses are actively amplified within cortex and contain an additional late component that is sensitive for context-specific sensory deviations. This strongly implicates deviance detection as a feature of intracortical stimulus processing beyond simple sensory input depression., Cerebral Cortex, 27 (1), ISSN:1047-3211, ISSN:1460-2199

Vibrissae-related sensorimotor cortex controls whisking movements indirectly via modulation of lower-level sensorimotor loops and a brainstem central pattern generator (CPG). Two different whisker representations in primary motor cortex (vM1) affect whisker movements in different ways. Prolonged microstimulation in RF, a larger anterior subregion of vM1, gives rise to complex face movements and whisker retraction while the same stimulation evokes large-amplitude rhythmic whisker movement in a small caudo-medial area (RW). To characterize the motor cortex representation of explorative whisking movements, here we recorded RW units in head-fixed rats trained to contact a moving object with one whisker. RW single units were found to encode two aspects of whisker movement independently, albeit on slow time scales (hundreds of milliseconds). The first is whisker position. The second consists of speed (absolute velocity), intensity (instantaneous power), and frequency (spectral centroid). The coding for the latter three parameters was tightly correlated and realized by a continuum of RW responses-ranging from a preference of movement to a preference of rest. Information theory analysis indicated that RW spikes carry most information about position and frequency, while intensity and speed are less well represented. Further, investigating multiple and single RW units, we found a lack of phase locking, movement anticipation, and contact-related tactile responses. These findings suggest that RW neither programs detailed whisker trajectories nor initiates them. Nor does it play a role in processing object touch. Its relationship to whisking is thus indirect and may be related to movement monitoring, perhaps using feedback from the CPG.

This paper describes experimental techniques with head-fixed, operantly conditioned rodents that allow the control of stimulus presentation and tracking of motor output at hitherto unprecedented levels of spatio-temporal precision. Experimental procedures for the surgery and behavioral training are presented. We place particular emphasis on potential pitfalls using these procedures in order to assist investigators who intend to engage in this type of experiment. We argue that head-fixed rodent models, by allowing the combination of methodologies from molecular manipulations, intracellular electrophysiology, and imaging to behavioral measurements, will be instrumental in combining insights into the functional neuronal organization at different levels of observation. Provided viable behavioral methods are implemented, model systems based on rodents will be complementary to current primate models--the latter providing highest comparability with the human brain, while the former offer hugely advanced methodologies on the lower levels of organization, for example, genetic alterations, intracellular electrophysiology, and imaging.

The present work introduces an electrode microdrive system small enough to be placed into two distinct brain areas in head-fixed mice. To meet the space constraint imposed by the size of mice and the additional presence of a head post, the size and weight of the components were minimized. In the current version, an individual microdrive moves an array of four Reitboeck type electrodes. We report about successful implantation in rats and mice using one or two microdrives. Using two of these devices in individual mice/rats, the recording of parallel single and multi-unit as well as local field potential from prefrontal, motor, somatosensory cortex and hippocampus is demonstrated. The system can be easily constructed with machinery and equipment present in most neurophysiology labs.

Little is known about the functional role of axotomized cortical neurons that survive spinal cord injury. Large thoracic spinal cord injuries in adult rats result in impairments of hindlimb function. Using retrograde tracers, we found that axotomized corticospinal axons from the hindlimb sensorimotor cortex sprouted in the cervical spinal cord. Mapping of these neurons revealed the emergence of a new forelimb corticospinal projection from the rostral part of the former hindlimb cortex. Voltage-sensitive dye (VSD) imaging and blood-oxygen-level-dependent functional magnetic resonance imaging (BOLD fMRI) revealed a stable expansion of the forelimb sensory map, covering in particular the former hindlimb cortex containing the rewired neurons. Therefore, axotomized hindlimb corticospinal neurons can be incorporated into the sensorimotor circuits of the unaffected forelimb.

Two-photon microscopy is a powerful method in biomedical research that allows functional and anatomical imaging at a subcellular resolution in vivo. The technique is seriously hampered by absorption and scattering of light by blood, which prevents imaging through large vessels. Here, we demonstrate in the rat cerebral cortex that blood replacement by perfluorocarbon emulsion, a compound also used in human critical care medicine, yields superior image quality, while preserving neuronal integrity. Shadows of large superficial vessels disappear completely and cells can be imaged underneath them. For the first time, it is possible to image complete populations of neurons and astrocytes in the upper layers of neocortex in vivo.

International audience; Tactile information is actively acquired and processed in the brain through concerted interactions between movement and sensation. Somatosensory input is often the result of self-generated movement during the active touch of objects, and conversely, sensory information is used to refine motor control. There must therefore be important interactions between sensory and motor pathways, which we chose to investigate in the mouse whisker sensorimotor system. Voltage-sensitive dye was applied to the neocortex of mice to directly image the membrane potential dynamics of sensorimotor cortex with subcolumnar spatial resolution and millisecond temporal precision. Single brief whisker deflections evoked highly distributed depolarizing cortical sensory responses, which began in the primary somatosensory barrel cortex and subsequently excited the whisker motor cortex. The spread of sensory information to motor cortex was dynamically regulated by behavior and correlated with the generation of sensory-evoked whisker movement. Sensory processing in motor cortex may therefore contribute significantly to active tactile sensory perception.

Biomedical optics express 6(11), 4228 (2015). doi:10.1364/BOE.6.004228, Published by OSA, Washington, DC

Neocortical responses typically adapt to repeated sensory stimulation, improving sensitivity to stimulus changes, but possibly also imposing limitations on perception. For example, it is unclear whether information about stimulus frequency is perturbed by adaptation or encoded by precise response timing. We addressed this question in rat barrel cortex by comparing performance in behavioral tasks with either whisker stimulation, which causes frequency-dependent adaptation, or optical activation of cortically expressed channelrhodopsin-2, which elicits non-adapting neural responses. Circumventing adaption by optical activation substantially improved cross-hemispheric discrimination of stimulus frequency. This improvement persisted when temporal precision of optically evoked spikes was reduced. We were able to replicate whisker-driven behavior only by applying adaptation rules mimicking sensory-evoked responses to optical stimuli. Conversely, in a change-detection task, animals performed better with whisker than optical stimulation. Our results directly demonstrate that sensory adaptation critically governs the perception of stimulus patterns, decreasing fidelity under steady-state conditions in favor of change detection.

Rats and mice receive a constant bilateral stream of tactile information with their large mystacial vibrissae when navigating in their environment. In a two-alternative forced choice paradigm (2-AFC), head-fixed rats and mice learned to discriminate vibrotactile frequencies applied simultaneously to individual whiskers on the left and right sides of the snout. Mice and rats discriminated 90-Hz pulsatile stimuli from pulsatile stimuli with lower repetition frequencies (10–80 Hz) but with identical kinematic properties in each pulse. Psychometric curves displayed an average perceptual threshold of 50.6-Hz and 53.0-Hz frequency difference corresponding to Weber fractions of 0.56 and 0.58 in mice and rats, respectively. Both species performed >400 trials a day (>200 trials per session, 2 sessions/day), with a peak performance of >90% correct responses. In general, rats and mice trained in the identical task showed comparable psychometric curves. Behavioral readouts, such as reaction times, learning rates, trial omissions, and impulsivity, were also very similar in the two species. Furthermore, whisking of the animals before stimulus presentation reduced task performance. This behavioral paradigm, combined with whisker position tracking, allows precise stimulus control in the 2-AFC task for head-fixed rodents. It is compatible with state-of-the-art neurophysiological recording techniques, such as electrophysiology and two-photon imaging, and therefore represents a valuable framework for neurophysiological investigations of perceptual decision-making.

Conventional decoding methods in neuroscience aim to predict discrete brain states from multivariate correlates of neural activity. This approach faces two important challenges. First, a small number of examples are typically represented by a much larger number of features, making it hard to select the few informative features that allow for accurate predictions. Second, accuracy estimates and information maps often remain descriptive and can be hard to interpret. In this paper, we propose a model-based decoding approach that addresses both challenges from a new angle. Our method involves (i) inverting a dynamic causal model of neurophysiological data in a trial-by-trial fashion; (ii) training and testing a discriminative classifier on a strongly reduced feature space derived from trial-wise estimates of the model parameters; and (iii) reconstructing the separating hyperplane. Since the approach is model-based, it provides a principled dimensionality reduction of the feature space; in addition, if the model is neurobiologically plausible, decoding results may offer a mechanistically meaningful interpretation. The proposed method can be used in conjunction with a variety of modelling approaches and brain data, and supports decoding of either trial or subject labels. Moreover, it can supplement evidence-based approaches for model-based decoding and enable structural model selection in cases where Bayesian model selection cannot be applied. Here, we illustrate its application using dynamic causal modelling (DCM) of electrophysiological recordings in rodents. We demonstrate that the approach achieves significant above-chance performance and, at the same time, allows for a neurobiological interpretation of the results.

A lateral hemisection injury of the cervical spinal cord results in Brown-Séquard syndrome in humans and rats. The hands/forelimbs on the injured side are rendered permanently impaired, but the legs/hindlimbs recover locomotor functions. This is accompanied by increased use of the forelimb on the uninjured side. Nothing is known about the cortical circuits that correspond to these behavioral adaptations. In this study, on adult rats with cervical spinal cord lateral hemisection lesions (at segment C3/4), we explored the sensory representation and corticospinal projection of the intact (ipsilesional) cortex. Using blood oxygenation level-dependent functional magnetic resonance imaging and voltage-sensitive dye (VSD) imaging, we found that the cortex develops an enhanced representation of the unimpaired forepaw by 12 weeks after injury. VSD imaging also revealed the cortical spatio-temporal dynamics in response to electrical stimulation of the ipsilateral forepaw or hindpaw. Interestingly, stimulation of the ipsilesional hindpaw at 12 weeks showed a distinct activation of the hindlimb area in the intact, ipsilateral cortex, probably via the injury-spared spinothalamic pathway. Anterograde tracing of corticospinal axons from the intact cortex showed sprouting to recross the midline, innervating the spinal segments below the injury in both cervical and lumbar segments. Retrograde tracing of these midline-crossing axons from the cervical spinal cord (at segment C6/7) revealed the formation of a new ipsilateral forelimb representation in the cortex. Our results demonstrate profound reorganizations of the intact sensory-motor cortex after unilateral spinal cord injury. These changes may contribute to the behavioral adaptations, notably for the recovery of the ipsilesional hindlimb.

β+-sensitive probes are useful tools for the measurement of radiotracer kinetics in small animals. They allow the cost-effective development of new PET tracers and offer the possibility to investigate a variety of cerebral processes. The study's main aim was the in vivo evaluation of a probe system for cerebral surface acquisitions. The detector system is a 0.2-mm thick scintillating disk of 3-mm diameter, positioned close to the cerebral surface. The study consists of 4 subparts: (1) simulation of the detection volume, (2) direct comparison with the classic intracortical beta probe regarding its capability to acquire kinetic data, (3) test of the ability to detect local tracer accumulations during infraorbital nerve (ION) electrostimulation and (4) demonstration of the feasibility to measure tracer kinetics in awake animals. Kinetic data acquired with 18F-fluorodeoxyglucose and 15O-H2O were fitted with standard compartment models. The surface probe measurements were in good agreement with those obtained using the intracortical scintillator. ION electrostimulation induced a marked increase in tracer accumulation adequately detected by the surface probe. In the head-fixed animal, a marked change in FDG kinetics was detected between the awake and anesthetized state. The novel surface probe system proved to be a valuable instrument for in vivo radiotracer studies of the cerebral cortex. Its main advantage is the absence of any tissue damage. In addition, serial acquisitions of tracer kinetics in the awake animal turned out to be feasible.

Laser speckle imaging (LSI) based on the speckle contrast analysis is a simple and robust technique for imaging of heterogeneous dynamics. LSI finds frequent application for dynamical mapping of cerebral blood flow, as it features high spatial and temporal resolution. However, the quantitative interpretation of the acquired data is not straightforward for the common case of a speckle field formed by both by moving and localized scatterers such as blood cells and bone or tissue. Here we present a novel processing scheme, we call dynamic laser speckle imaging (dLSI), that can be used to correctly extract the temporal correlation parameters from the speckle contrast measured in the presence of a static or slow-evolving background. The static light contribution is derived from the measurements by cross-correlating sequential speckle images. In-vivo speckle imaging experiments performed in the rodent brain demonstrate that dLSI leads to improved results. The cerebral hemodynamic response observed through the thinned and intact skull are more pronounced in the dLSI images as compared to the standard speckle contrast analysis. The proposed method also yields benefits with respect to the quality of the speckle images by suppressing contributions of non-uniformly distributed specular reflections.

Stimulation mapping of motor cortex is an important tool for assessing motor cortex physiology. Existing techniques include intracortical microstimulation (ICMS) which has high spatial resolution but damages cortical integrity by needle penetrations, and transcranial stimulation which is non-invasive but lacks focality and spatial resolution. A minimally invasive epidural microstimulation (EMS) technique using chronically implanted polyimide-based thin-film microelectrode arrays (72 contacts) was tested in rat motor cortex and compared to ICMS within individual animals. Results demonstrate reliable mapping with high reproducibility and validity with respect to ICMS. No histological evidence of cortical damage and the absence of motor deficits as determined by performance of a motor skill reaching task, demonstrate the safety of the method. EMS is specifically suitable for experiments integrating electrophysiology with behavioral and molecular biology techniques.

Palpatory movements (‘active’ touch) are an integral part of tactile sensing. It is known that tactile signals can be modulated in certain behavioral contexts, but it is still unresolved to what degree this modulation is related to movement kinematics and whether it stems from tactile receptors or from central sources. Using awake, headfixed rats, trained to contact an object, we measured trajectories of muscle-propelled whisker movement precisely and compared tactile responses to contacts thus accomplished with ‘passive’ contacts (motionless whisker contacted by object). Multielectrode extracellular recordings in deep layers of barrel cortex revealed that when the animals moved their whiskers actively, tactile processing switched from high response amplitudes, wide cortical representation and low background firing, to low response amplitudes, narrow spatial representation and elevated background firing. Switching was fast (

What is mapped on the surface of the primary motor cortex (M1)? The classic somatotopic map holds true on the level of limb representations. However, on the small scale (at within-limb representations), neither somatotopy nor movement dynamics/kinematics seem to be organizational principles. We investigated the hypothesis that integrated into the body representation of M1 there may be separate representation of different modes of motor control, using different subcortical computations but sharing the same motor periphery. Using awake rats and long intracortical stimulation trains in M1 whisker representation (wM1) revealed that natural-like, rhythmic whisking (normally used for tactile exploration) can be evoked from a posteromedial subregion of wM1. Nonrhythmic whisker retraction, on the other hand, was evoked in an adjacent but more anterolaterally located region within wM1. Evoked whisker retraction was always accompanied by complex movements of the face, suggesting that the respective subregion is able to interact with other representations in specific behavioral contexts. Such associations were absent for evoked rhythmic whisking. The respective subregion rather seemed to activate a downstream central pattern generator, the oscillation frequency of which was dependent on the average evoked cortical activity. Nevertheless, joint stimulation of the two neighboring subregions demonstrated their potency to interact in a functionally useful way. Therefore, we suggest that the cause of cortical separation is the specific drive of subcortical structures needed to generate different types of movements rather than different behavioral contexts in which the movements are performed.

The experiments strongly suggested that the reason why Purkinje cells die so easily after global brain ischemia relates to deficiencies in aldolase C and EAAT4 that allow them to survive pathologically intense synaptic input from the inferior olive after the restoration of blood flow. This conclusion is based on: (a) the remarkably tight correspondence between the regional absence of aldolase C and EAAT4 in Purkinje cells and the patterned loss of Purkinje cells after a bout of global brain ischemia; (b) the necessity of the olivocerebellar pathway for the ischemic death of Purkinje cells; and (c) the build-up of pathologically synchronous and high-frequency burst activity within the inferior olive during recovery from ischemia. Indeed, the correspondence between the absence of aldolase C and EAAT4 to sensitivity to ischemia could be demonstrated for zones of Purkinje cells as small as two neurons. A second finding was that Purkinje cells are not uniformly sensitive to transient ischemia, since they die most frequently in zones where aldolase C and EAAT4 are absent. One implication of the experiment is that factors beyond the unique synaptic and membrane properties of Purkinje cells play an important role in determining this neuron's high sensitivity to ischemia. The data strongly imply that two properties of Purkinje cells that make them susceptible to ischemic death are their reduced capability to sequester glutamate and reduced ability to generate energy during anoxia. The patterned death of Purkinje cells is sufficient to induce a form of audiogenic myoclonus, as determined with a neurotoxic dose of ibogaine. Ibogaine-induced myoclonus is recognized behaviorally as a reduced ability to habituate to a startle stimulus and resembles the myoclonic jerk of rats during recovery from a prolonged bout of global brain ischemia. Commonalities of ischemia and ibogaine-induced neurodegeneration are the intricately striped Purkinje cell loss in the posterior lobe and a nearly complete deafferentation of the lateral aspect of the fastigial nucleus from the cerebellar cortex, in particular the dorsolateral protuberance. Thus, the data point strongly to a cerebellar contribution to audiogenic myoclonus. Single-neuron electrophysiology experiments in monkeys have demonstrated that the evoked activity in the deep cerebellar nuclei occurs too late to initiate the startle response (60) and electromyography of the postischemic myoclonus of rats corroborates this view (see Chapter 31) (20). However, the nearly complete loss of GABAergic terminals in the dorsolateral protuberance after Purkinje cell death would be expected to dramatically increase its tonic firing and the background excitation of the brain-stem structures that it innervates. The fastigial nucleus innervates a large number of autonomic and motor structures in the brainstem and diencephalon, including the ventrolateral nucleus of the thalamus and the gigantocellular reticular nucleus in the medulla--structures that have been implicated in human posthypoxic myoclonus (6, 7). We propose that the posthypoxic myoclonic jerk of rats is, at least in part, due to disinhibition of the fastigial nucleus produced by patterned Purkinje cell death in the vermis. The argument is as follows: the loss of GABAergic inhibition in the fastigial nucleus after ischemia leads to diaschisis of the motor thalamus and reticular formation which, in turn, is responsible for enhanced motor excitability and myoclonus. That the audiogenic myoclonus after global brain ischemia in the rat gradually resolves over a period of 2 to 3 weeks is consistent with this view, as restoration of background excitability after CNS damage in rats has been documented to occur within this time-frame (61). Our view brings together the physiologic finding that posthypoxic myoclonus appears to originate in the sensory-motor cortices and/or reticular formation with the consistent anatomical finding of Purkinje cell loss after ischemia, and explains the puzzle of Marsden's unique cases of myoclonus associated with coeliac disease (1). Moreover, our argument is consistent with findings both in rats (62, 63) and humans (64) that damage to the vermis impairs the long-term habituation of the startle reflex. It remains to be determined whether the pathologically enhanced startle responses after vermal damage resemble brain-stem reticular or cortical myoclonus at the electrophysiologic level of analysis. What is the purpose of the regional expression of aldolase C and EAAT4 in Purkinje cells? The close correspondence between the spatial distribution of aldolase C and the parasagittal anatomy of the cerebellum (48) has led to the view that aldolase C may help specify connectivity during development. While the present experiments do not address this issue, they underscore the fact that aldolase plays a fundamental role in metabolism. Because Purkinje cells have a repressed expression of aldolase A (31), whatever role the absence of aldolase C may play during development comes at the price of metabolic frailty later in adulthood. From another point of view, aldolase C and EAAT4 appear to confer upon Purkinje cells the ability to survive their own climbing fiber. Indeed, climbing fibers form a distributed synapse that synchronously releases glutamate (or aspartate) at all levels of the dendritic tree simultaneously (65, 66). Such synchronous activation triggers calcium influx throughout the Purkinje cell dendrites at a magnitude that is unparalleled in the nervous system (12), and, thus, places an extraordinarily high metabolic demand on the Purkinje cell. The apparently reduced level of aldolase in a subpopulation of Purkinje cells provides the condition for energy failure and death during anoxia so long as the climbing fibers are intact or when climbing fiber activation is pharmacologically enhanced under normoxic conditions, such as after ibogaine (53-56). Lastly, the argument that diaschisis produced by patterned cerebellar degeneration leads to thalamo-cortical and reticular hyperexcitability agrees with C. David Marsden and his colleagues' bold demonstration of an inhibitory influence of cerebellar cortex on motor cortex in humans (67). Our anatomic data indicate that the spatially distinct zones of Purkinje cells, which are killed by global brain ischemia, may be the origin of such inhibition.

Sensory maps are reshaped by experience. It is unknown how map plasticity occurs in vivo in functionally diverse neuronal populations because activity of the same cells has not been tracked over long time periods. Here we used repeated two-photon imaging of a genetic calcium indicator to measure whisker-evoked responsiveness of the same layer 2/3 neurons in adult mouse barrel cortex over weeks, first with whiskers intact, then during continued trimming of all but one whisker. Across the baseline period, neurons displayed heterogeneous yet stable responsiveness. During sensory deprivation, responses to trimmed whisker stimulation globally decreased, whereas responses to spared whisker stimulation increased for the least active neurons and decreased for the most active neurons. These findings suggest that recruitment of inactive, 'silent' neurons is part of a convergent redistribution of population activity underlying sensory map plasticity. Sensory-driven responsiveness is a key property controlling experience-dependent activity changes in individual neurons.

International audience; Virtual reality (VR) has recently become an affordable technology. A wide range of options are available to access this unique visualization medium, from simple cardboard inserts for smartphones to truly advanced headsets tracked by external sensors. While it is now possible for any research team to gain access to VR, we can still question what it brings to scientific research. Visualization and the ability to navigate complex three-dimensional data are undoubtedly a gateway to many scientific applications; however, we are convinced that data treatment and numerical simulations, especially those mixing interactions with data, human cognition, and automated algorithms will be the future of VR in scientific research. Moreover, VR might soon merit the same level of attention to imaging data as machine learning currently has. In this short perspective, we discuss approaches that employ VR in scientific research based on some concrete examples.