The dorsal striatum is crucial for the acquisition and consolidation of instrumental behaviour, but the underlying computations and internal dynamics remain elusive. To address this issue, we combined a model of key computations supporting decision-making during instrumental learning with human behavioural and functional magnetic resonance imaging (fMRI) data. The results showed that the associative and sensorimotor dorsal striatum host complementary computations that, we suggest, may differentially support goal-directed and habitual processes. The anterior caudate nucleus integrates information about performance and cognitive control demands, whereas the putamen tracks how likely the conditioning stimuli lead to correct response. Contrary to current models, the putamen is recruited during initial acquisition. As the exploratory phase proceeds, the relative contribution of the caudate nucleus becomes dominant over the putamen. During early consolidation, caudate nucleus and putamen settle to asymptotic values and share control. We then investigated how dorsal striatal computations may affect decision-making. We found that portion of reaction times' variance parallels the combined cost associated with the dorsal striatal computations. Overall, our findings provide a deeper insight into the functional heterogeneity within the dorsal striatum and suggest that the dynamic interplay between caudate nucleus and putamen, rather than their serial recruitment, underlies the acquisition and early consolidation of instrumental behaviours.

While learning from an instructor by watching a 'how-to' video has become common practice, we know surprisingly little about the relation between brain activities in instructor and observers. In this fMRI study we investigated the temporal synchronization between instructor and observers using intersubject correlation in the naturalistic setting of learning to fold origami. Brain activity of the blindfolded instructor during action production was compared to the observers while they viewed the instructor's video-taped actions. We demonstrate for the first time that the BOLD activity in the instructor's and observer's brain are synchronized while observing and learning a manual complex task with the goal of reproducing it. We can rule out that this synchrony originates from visual feedback. Observers exhibiting higher synchrony with the instructor in the ventral premotor cortex, while viewing the video for the first time, were more successful in reproducing the origami afterwards. Furthermore, changes in instructor-observer synchrony across observational learning sessions occur in cerebellar areas, as well as differences in instructor-observer synchrony between learning and the counting folds, our non-learning control. Not only known cerebellar motor production areas show synchrony, shedding new light on the involvement of the cerebellum in action observation and learning.

Theoretical models and experimental evidence suggest that cortico-striatal-cerebellar networks play a crucial role in mediating motor sequence learning. However, how these different regions interact in order to mediate learning is less clear. In the present fMRI study, we used dynamic causal modeling to investigate effective connectivity within the cortico-striatal-cerebellar network while subjects performed a serial reaction time task. Using Bayesian model selection and family wise inference, we show that the cortico-cerebellar loop had higher model evidence than the cortico-striatal loop during motor learning. We observed significant negative modulatory effects on the connections from M1 to cerebellum bilaterally during learning. The results suggest that M1 causes the observed decrease in activity in the cerebellum as learning progresses. The current study stresses the significant role that the cerebellum plays in motor learning as previously suggested by fMRI studies in healthy subjects as well as behavioral studies in patients with cerebellar dysfunction. These results provide important insight into the neural mechanisms underlying motor learning.

Neuroimaging studies have improved our understanding of which brain structures are involved in motor learning. Despite this, questions remain regarding the areas that contribute consistently across paradigms with different task demands. For instance, sensorimotor tasks focus on learning novel movement kinematics and dynamics, while serial response time task (SRTT) variants focus on sequence learning. These differing task demands are likely to elicit quantifiably different patterns of neural activity on top of a potentially consistent core network. The current study identified consistent activations across 70 motor learning experiments using activation likelihood estimation (ALE) meta-analysis. A global analysis of all tasks revealed a bilateral cortical-subcortical network consistently underlying motor learning across tasks. Converging activations were revealed in the dorsal premotor cortex, supplementary motor cortex, primary motor cortex, primary somatosensory cortex, superior parietal lobule, thalamus, putamen and cerebellum. These activations were broadly consistent across task specific analyses that separated sensorimotor tasks and SRTT variants. Contrast analysis indicated that activity in the basal ganglia and cerebellum was significantly stronger for sensorimotor tasks, while activity in cortical structures and the thalamus was significantly stronger for SRTT variants. Additional conjunction analyses then indicated that the left dorsal premotor cortex was activated across all analyses considered, even when controlling for potential motor confounds. The highly consistent activation of the left dorsal premotor cortex suggests it is a critical node in the motor learning network.

In conventional non-quantitative magnetic resonance imaging, image contrast is consistent within images, but absolute intensity can vary arbitrarily between scans. For quantitative analysis of intensity data, images are typically normalized to a consistent reference. The most convenient reference is a tissue that is always present in the image, and is unlikely to be affected by pathological processes. In multiple sclerosis neuroimaging, both the white and gray matter are affected, so normalization techniques that depend on brain tissue may introduce bias or remove biological changes of interest. We introduce a complementary procedure, image "calibration," the goal of which is to remove technical intensity artifacts while preserving biological differences. We demonstrate a deep learning approach to segmenting fat from within the orbit of the eyes on T1-weighted images at 1.5 and 3 ​T to use as a reference tissue, and use it to calibrate 1018 scans from 256 participants in a study of pediatric-onset multiple sclerosis. The machine segmentations agreed with the adjudicating expert (DF) segmentations better than did those of other expert humans, and calibration resulted in better agreement with semi-quantitative magnetization transfer ratio imaging than did normalization with the WhiteStripe1 algorithm. We suggest that our method addresses two key priorities in the field: (1) it provides a robust option for serial calibration of conventional scans, allowing comparison of disease change in persons imaged at multiple time points in their disease; and (ii) the technique is fast, as the deep learning segmentation takes only 0.5 ​s/scan, which is feasible for both large and small datasets.

Volumetric magnetic resonance imaging studies have shown that intense learning can be associated with grey matter volume increases in the adult brain. The underlying mechanisms are poorly understood. Here we used monocular deprivation in rats to analyze the mechanisms underlying use-dependent grey matter increases. Optometry for quantification of visual acuity was combined with volumetric magnetic resonance imaging and microscopic techniques in longitudinal and cross-sectional studies. We found an increased spatial vision of the open eye which was associated with a transient increase in the volumes of the contralateral visual and lateral entorhinal cortex. In these brain areas dendrites of neurons elongated, and there was a strong increase in the number of spines, the targets of synapses, which was followed by spine maturation and partial pruning. Astrocytes displayed a transient pronounced swelling and underwent a reorganization of their processes. The use-dependent increase in grey matter corresponded predominantly to the swelling of the astrocytes. Experience-dependent increase in brain grey matter volume indicates a gain of structure plasticity with both synaptic and astrocyte remodeling.

Transcranial direct current stimulation (tDCS) is a non-invasive brain stimulation approach in which low level currents are administered over the scalp to influence underlying brain function. Prevailing theories of tDCS focus on modulation of excitation-inhibition balance at the local stimulation location. However, network level effects are reported as well, and appear to depend upon differential underlying mechanisms. Here, we evaluated potential network-level effects of tDCS during the Serial Reaction Time Task (SRTT) using convergent EEG- and fMRI-based connectivity approaches. Motor learning manifested as a significant (p<.0001) shift from slow to fast responses and corresponded to a significant increase in beta-coherence (p<.0001) and fMRI connectivity (p<.01) particularly within the visual-motor pathway. Differential patterns of tDCS effect were observed within different parametric task versions, consistent with network models. Overall, these findings demonstrate objective physiological effects of tDCS at the network level that result in effective behavioral modulation when tDCS parameters are matched to network-level requirements of the underlying task.

Alpha oscillations (8-13 Hz) have been suggested to play an important role in dynamic neural processes underlying learning and memory. The goal of this study was to scrutinize the role of alpha oscillations in communication within a cortico-cerebellar network implicated in motor sequence learning. To this end, we conducted two EEG experiments using a serial reaction time task. In the first experiment, we explored changes in alpha power and cross-channel alpha coherence as subjects learned a motor sequence. We found a gradual decrease in spectral alpha power over left premotor cortex (PMC) and sensorimotor cortex (SM1) during learning blocks. In addition, alpha coherence between left PMC/SM1 and left cerebellar crus I was specifically decreased during sequence learning, possibly reflecting a functional decoupling in the broader motor learning network. In the second experiment in a different cohort, we applied 10Hz transcranial alternating current stimulation (tACS), a method shown to entrain local oscillatory activity, to left M1 (lM1) and right cerebellum (rCB) during sequence learning. We observed a tendency for diminished learning following rCB tACS compared to sham, but not following lM1 tACS. Learning-related alpha power following rCB tACS was increased in left PMC, possibly reflecting increase in local inhibitory neural activity. Importantly, learning-specific alpha coherence between left PMC and right cerebellar lobule VIIb was enhanced following rCB tACS. These findings provide strong evidence for a causal role of alpha oscillations in controlling information transfer in a premotor-cerebellar loop during motor sequence learning. Our findings are consistent with a model in which sequence learning may be impaired by enhancing premotor cortical alpha oscillation via external modulation of cerebellar oscillations.

This magnetoencephalography (MEG) study investigates how procedural sequence learning performance is related to prior brain resting-state functional connectivity (rsFC), and to what extent sequence learning induces rapid changes in brain rsFC in school-aged children. Procedural learning was assessed in 30 typically developing children (mean age ± SD: 9.99 years ± 1.35) using a serial reaction time task (SRTT). During SRTT, participants touched as quickly and accurately as possible a stimulus sequentially or randomly appearing in one of the quadrants of a touchscreen. Band-limited power envelope correlation (brain rsFC) was applied to MEG data acquired at rest pre- and post-learning. Correlation analyses were performed between brain rsFC and sequence-specific learning or response time indices. Stronger pre-learning interhemispheric rsFC between inferior parietal and primary somatosensory/motor areas correlated with better subsequent sequence learning performance and faster visuomotor response time. Faster response time was associated with post-learning decreased rsFC within the dorsal extra-striate visual stream and increased rsFC between temporo-cerebellar regions. In school-aged children, variations in functional brain architecture at rest within the sensorimotor network account for interindividual differences in sequence learning and visuomotor performance. After learning, rapid adjustments in functional brain architecture are associated with visuomotor performance but not sequence learning skills.

Leading neuroimaging studies have pushed 3T MRI acquisition resolutions below 1.0 mm for improved structure definition and morphometry. Yet, only few, time-intensive automated image analysis pipelines have been validated for high-resolution (HiRes) settings. Efficient deep learning approaches, on the other hand, rarely support more than one fixed resolution (usually 1.0 mm). Furthermore, the lack of a standard submillimeter resolution as well as limited availability of diverse HiRes data with sufficient coverage of scanner, age, diseases, or genetic variance poses additional, unsolved challenges for training HiRes networks. Incorporating resolution-independence into deep learning-based segmentation, i.e., the ability to segment images at their native resolution across a range of different voxel sizes, promises to overcome these challenges, yet no such approach currently exists. We now fill this gap by introducing a Voxel-size Independent Neural Network (VINN) for resolution-independent segmentation tasks and present FastSurferVINN, which (i) establishes and implements resolution-independence for deep learning as the first method simultaneously supporting 0.7-1.0 mm whole brain segmentation, (ii) significantly outperforms state-of-the-art methods across resolutions, and (iii) mitigates the data imbalance problem present in HiRes datasets. Overall, internal resolution-independence mutually benefits both HiRes and 1.0 mm MRI segmentation. With our rigorously validated FastSurferVINN we distribute a rapid tool for morphometric neuroimage analysis. The VINN architecture, furthermore, represents an efficient resolution-independent segmentation method for wider application.

Changes in resting state functional connectivity are induced by sensorimotor training and assumed to be concomitant of motor learning, although a potential relationship between functional and structural connectivity associated with sensorimotor sequence learning remains elusive. To investigate whether initial structural connectivity relates to changes in functional connectivity, we evaluated resting state functional connectivity (rs-FC), white matter fibre density (FD), fibre-bundle cross-section (FC), and gray matter volume (GMV) in healthy human participants before and after two days of performing a complex whole-body serial reaction time task (CWB-SRTT). As CWB-SRTT was implicit, participants were not told about the presence of any sequence. Since the lateral prefrontal cortex (PFC) plays an important role in sequence learning, we hypothesized that structural connectivity within the PFC prior to learning is associated with changes in rs-FC involving the lateral PFC. Sequence specific improvements, as assessed by the time difference between the last random and the last sequence blocks, were observed for reaction times, suggesting that sensorimotor sequence memory was acquired. Rs-FC between the right lateral PFC and bilateral striatum increased significantly in the learning group, when compared to a control group who performed only random blocks. This indicated that rs-FC changes are related to sequence memory but not to exercise itself. In addition, changes in rs-FC between the right lateral PFC and the left striatum were correlated with sequence specific improvements in individual reaction times. Furthermore, changes in rs-FC between right lateral PFC and left striatum were positively correlated with FC in the right anterior corona radiata measured before the task. We did not find any structural changes or significant correlations in FD or GMV. These findings suggest that an early phase of sensorimotor sequence learning in complex whole-body movements is associated with an increase in rs-FC between prefrontal and subcortical regions. Furthermore, we provide novel evidence that CWB-SRTT-induced changes in rs-FC were correlated with FC within the PFC.

The fast and slow learning stages of motor sequence learning are suggested to be realized through plasticity in a distributed cortico-striato-cerebellar network. To better understand the causal interactions within this network in the different phases of motor sequence learning, we investigated the effective connectivity within this network during encoding (Day 1) and after consolidation (Day 2) of a serial reaction time task. Using Dynamic Causal Modelling of fMRI data, we found general changes in network connections reflected in altered input nodes and endogenous connections when comparing the early and fast learning session to the late and slow learning session. Whereas encoding of a motor memory early on modulated several connections in a distributed network, slow learning resulted in a pruned network. More specifically, we found a negative modulation of connections from left M1 to right cerebellum, right premotor cortex to left cerebellum, as well as backward connections from putamen to cerebellum bilaterally in the encoding session. While connections during pre-sleep were significantly modulated by learning per se (i.e., specifically modulated by performance on sequence conditions), the connections observed after sleep were rather modulated by general performance (i.e., modulated by performance on both sequence and random conditions). A forward connection from left cerebellum to right putamen was found to be consistent across participants for the sequence condition only during slow learning. Together these findings suggest that whereas encoding in the fast learning phase requires plasticity in several connections implementing both motor and perceptual learning components, slow learning is mediated through connectivity from left cerebellum to right putamen.

Non-invasive transcranial stimulation of cerebellum and primary motor cortex (M1) has been shown to enhance motor learning. However, the mechanisms by which stimulation improves learning remain largely unknown. Here, we sought to shed light on the neural correlates of transcranial direct current stimulation (tDCS) during motor learning by simultaneously recording functional magnetic resonance imaging (fMRI). We found that right cerebellar tDCS, but not left M1 tDCS, led to enhanced sequence learning in the serial reaction time task. Performance was also improved following cerebellar tDCS compared to sham in a sequence production task, reflecting superior training effects persisting into the post-training period. These behavioral effects were accompanied by increased learning-specific activity in right M1, left cerebellum lobule VI, left inferior frontal gyrus and right inferior parietal lobule during cerebellar tDCS compared to sham. Despite the lack of group-level changes comparing left M1 tDCS to sham, activity increase in right M1, supplementary motor area, and bilateral middle frontal cortex, under M1 tDCS, was associated with better sequence performance. This suggests that lack of group effects in M1 tDCS relate to inter-individual variability in learning-related activation patterns. We further investigated how tDCS modulates effective connectivity in the cortico-striato-cerebellar learning network. Using dynamic causal modelling, we found altered connectivity patterns during both M1 and cerebellar tDCS when compared to sham. Specifically, during cerebellar tDCS, negative modulation of a connection from putamen to cerebellum was decreased for sequence learning only, effectively leading to decreased inhibition of the cerebellum. These results show specific effects of cerebellar tDCS on functional activity and connectivity in the motor learning network and may facilitate the optimization of motor rehabilitation involving cerebellar non-invasive stimulation.

Short-term memory (STM) for order information, as compared to STM for item information, has been shown to be a critical determinant of language learning capacity. The present fMRI study asked whether the neural substrates of order STM can serve as markers for bilingual language achievement. Two groups of German-French bilinguals differing in second language proficiency were presented STM tasks probing serial order or item information. During order STM but not item STM tasks, the high proficiency group showed increased activation in the lateral orbito-frontal and the superior frontal gyri associated with updating and grouped rehearsal of serial order information. Functional connectivity analyses for order encoding showed a functional network involving the left IPS, the right IPS and the right superior cerebellum in the high proficiency group while the low proficiency group showed enhanced connectivity between the left IPS and bilateral superior temporal and temporo-parietal areas involved in item processing. The present data suggest that low proficiency bilinguals activate STM networks for order in a less efficient and differentiated way, and this may explain their poorer storage and learning capacity for verbal sequences.

Age differences in the strategies that individuals spontaneously use to learn new information have been shown to contribute to age differences in episodic memory. We investigated the role of prefrontal structure in observed age effects on self-initiated use of memory strategies. The relationships among age, prefrontal regional gray matter volumes, and semantic and serial clustering during free recall on the California Verbal Learning Test-II were examined across the adult lifespan. Semantic clustering was negatively correlated with age and positively correlated with gray matter volumes in bilateral middle and left inferior frontal regions across the adult lifespan. Gray matter volumes in these regions mediated the effects of age on semantic clustering. Forward serial clustering was also negatively correlated with age. However, forward serial clustering was not significantly positively correlated with gray matter volumes in any region of lateral prefrontal cortex. These results suggest that bilateral middle and left inferior frontal regions support self-initiated semantic memory strategy use across the adult lifespan. They also suggest that age differences in prefrontal gray matter volume are a significant contributor to age differences in self-initiated use of elaborative memory strategies.

The human mediodorsal thalamic nucleus (MD) is crucial for higher cognitive functions, while the fine anatomical organization of the MD and the function of each subregion remain elusive. In this study, using high-resolution data provided by the Human Connectome Project, an anatomical connectivity-based method was adopted to unveil the topographic organization of the MD. Four fine-grained subregions were identified in each hemisphere, including the medial (MDm), central (MDc), dorsal (MDd), and lateral (MDl), which recapitulated previous cytoarchitectonic boundaries from histological studies. The subsequent connectivity analysis of the subregions also demonstrated distinct anatomical and functional connectivity patterns, especially with the prefrontal cortex. To further evaluate the function of MD subregions, partial least squares analysis was performed to examine the relationship between different prefrontal-subregion connectivity and behavioral measures in 1012 subjects. The results showed subregion-specific involvement in a range of cognitive functions. Specifically, the MDm predominantly subserved emotional-cognition domains, while the MDl was involved in multiple cognitive functions especially cognitive flexibility and inhibition. The MDc and MDd were correlated with fluid intelligence, processing speed, and emotional cognition. In conclusion, our work provides new insights into the anatomical and functional organization of the MD and highlights the various roles of the prefrontal-thalamic circuitry in human cognition.

The surface of the human cerebellar cortex is much more tightly folded than the cerebral cortex. Volumetric analysis of cerebellar morphometry in magnetic resonance imaging studies suffers from insufficient resolution, and therefore has had limited impact on disease assessment. Automatic serial polarization-sensitive optical coherence tomography (as-PSOCT) is an emerging technique that offers the advantages of microscopic resolution and volumetric reconstruction of large-scale samples. In this study, we reconstructed multiple cubic centimeters of ex vivo human cerebellum tissue using as-PSOCT. The morphometric and optical properties of the cerebellar cortex across five subjects were quantified. While the molecular and granular layers exhibited similar mean thickness in the five subjects, the thickness varied greatly in the granular layer within subjects. Layer-specific optical property remained homogenous within individual subjects but showed higher cross-subject variability than layer thickness. High-resolution volumetric morphometry and optical property maps of human cerebellar cortex revealed by as-PSOCT have great potential to advance our understanding of cerebellar function and diseases.

We have previously shown that several brain areas are activated both during sequence learning at wake and during subsequent rapid-eye-movements (REM) sleep (Nat. Neurosci. 3 (2000) 831-836), suggesting that REM sleep participates in the reprocessing of recent memory traces in humans. However, the nature of the reprocessed information remains open. Here, we show that regional cerebral reactivation during posttraining REM sleep is not merely related to the acquisition of basic visuomotor skills during prior practice of the serial reaction time task, but rather to the implicit acquisition of the probabilistic rules that defined stimulus sequences. Moreover, functional connections between the reactivated cuneus and the striatum--the latter being critical for implicit sequence learning--are reinforced during REM sleep after practice on a probabilistic rather than on a random sequence of stimuli. Our results therefore support the hypothesis that REM sleep is deeply involved in the reprocessing and optimization of the high-order information contained in the material to be learned. In addition, we show that the level of acquisition of probabilistic rules attained prior to sleep is correlated to the increase in regional cerebral blood flow during subsequent REM sleep. This suggests that posttraining cerebral reactivation is modulated by the strength of the memory traces developed during the learning episode. Our data provide the first experimental evidence for a link between behavioral performance and cerebral reactivation during REM sleep.

Measures of change in hippocampal volume derived from longitudinal MRI are a well-studied biomarker of disease progression in Alzheimer's disease (AD) and are used in clinical trials to track therapeutic efficacy of disease-modifying treatments. However, longitudinal MRI change measures based on deformable registration can be confounded by MRI artifacts, resulting in over-estimation or underestimation of hippocampal atrophy. For example, the deformation-based-morphometry method ALOHA (Das et al., 2012) finds an increase in hippocampal volume in a substantial proportion of longitudinal scan pairs from the Alzheimer's Disease Neuroimaging Initiative (ADNI) study, unexpected, given that the hippocampal gray matter is lost with age and disease progression. We propose an alternative approach to quantify disease progression in the hippocampal region: to train a deep learning network (called DeepAtrophy) to infer temporal information from longitudinal scan pairs. The underlying assumption is that by learning to derive time-related information from scan pairs, the network implicitly learns to detect progressive changes that are related to aging and disease progression. Our network is trained using two categorical loss functions: one that measures the network's ability to correctly order two scans from the same subject, input in arbitrary order; and another that measures the ability to correctly infer the ratio of inter-scan intervals between two pairs of same-subject input scans. When applied to longitudinal MRI scan pairs from subjects unseen during training, DeepAtrophy achieves greater accuracy in scan temporal ordering and interscan interval inference tasks than ALOHA (88.5% vs. 75.5% and 81.1% vs. 75.0%, respectively). A scalar measure of time-related change in a subject level derived from DeepAtrophy is then examined as a biomarker of disease progression in the context of AD clinical trials. We find that this measure performs on par with ALOHA in discriminating groups of individuals at different stages of the AD continuum. Overall, our results suggest that using deep learning to infer temporal information from longitudinal MRI of the hippocampal region has good potential as a biomarker of disease progression, and hints that combining this approach with conventional deformation-based morphometry algorithms may lead to improved biomarkers in the future.

The ability to walk independently with the velocity and endurance that permit home and community activities is a highly regarded goal for neurological rehabilitation after stroke. This pilot study explored a functional magnetic resonance imaging (fMRI) activation paradigm for its ability to reflect phases of motor learning over the course of locomotor rehabilitation-mediated functional gains. Ankle dorsiflexion is an important kinematic aspect of the swing and initial stance phase of the gait cycle. The motor control of dorsiflexion depends in part on descending input from primary motor cortex. Thus, an fMRI activation paradigm using voluntary ankle dorsiflexion has face validity for the serial study of walking-related interventions. Healthy control subjects consistently engaged contralateral primary sensorimotor cortex (S1M1), supplementary motor area (SMA), premotor (PM) and cingulate motor (CMA) cortices, and ipsilateral cerebellum. Four adults with chronic hemiparetic stroke evolved practice-induced representational plasticity associated with gains in speed, endurance, motor control, and kinematics for walking. For example, an initial increase in activation within the thoracolumbar muscle representation of S1M1 in these subjects was followed by more focused activity toward the foot representation with additional pulses of training. Contralateral CMA and the secondary sensory area also reflected change with practice and gains. We demonstrate that the supraspinal sensorimotor network for the neural control of walking can be assessed indirectly by ankle dorsiflexion. The ankle paradigm may serve as an ongoing physiological assay of the optimal type, duration, and intensity of rehabilitative gait training.