Generation of synthetic EEG data for training algorithms supporting the diagnosis of major depressive disorder.

Introduction: Major depressive disorder (MDD) is the most common mental disorder worldwide, leading to impairment in quality and independence of life. Electroencephalography (EEG) biomarkers processed with machine learning (ML) algorithms have been explored for objective diagnoses with promising results. However, the generalizability of those models, a prerequisite for clinical application, is restricted by small datasets. One approach to train ML models with good generalizability is complementing the original with synthetic data produced by generative algorithms. Another advantage of synthetic data is the possibility of publishing the data for other researchers without risking patient data privacy. Synthetic EEG time-series have not yet been generated for two clinical populations like MDD patients and healthy controls. Methods: We first reviewed 27 studies presenting EEG data augmentation with generative algorithms for classification tasks, like diagnosis, for the possibilities and shortcomings of recent methods. The subsequent empirical study generated EEG time-series based on two public datasets with 30/28 and 24/29 subjects (MDD/controls). To obtain baseline diagnostic accuracies, convolutional neural networks (CNN) were trained with time-series from each dataset. The data were synthesized with generative adversarial networks (GAN) consisting of CNNs. We evaluated the synthetic data qualitatively and quantitatively and finally used it for re-training the diagnostic model. Results: The reviewed studies improved their classification accuracies by between 1 and 40% with the synthetic data. Our own diagnostic accuracy improved up to 10% for one dataset but not significantly for the other. We found a rich repertoire of generative models in the reviewed literature, solving various technical issues. A major shortcoming in the field is the lack of meaningful evaluation metrics for synthetic data. The few studies analyzing the data in the frequency domain, including our own, show that only some features can be produced truthfully. Discussion: The systematic review combined with our own investigation provides an overview of the available methods for generating EEG data for a classification task, their possibilities, and shortcomings. The approach is promising and the technical basis is set. For a broad application of these techniques in neuroscience research or clinical application, the methods need fine-tuning facilitated by domain expertise in (clinical) EEG research.

Two Routes to Alzheimer's Disease Based on Differential Structural Changes in Key Brain Regions.

BACKGROUND: Alzheimer's disease (AD) is a neurodegenerative disorder with homogenous disease patterns. Neuropathological changes precede symptoms by up to two decades making neuroimaging biomarkers a prime candidate for early diagnosis, prognosis, and patient stratification. OBJECTIVE: The goal of the study was to discern intermediate AD stages and their precursors based on neuroanatomical features for stratifying patients on their progression through different stages. METHODS: Data include grey matter features from 14 brain regions extracted from longitudinal structural MRI and cognitive data obtained from 1,017 healthy controls and AD patients of ADNI. AD progression was modeled with a Hidden Markov Model, whose hidden states signify disease stages derived from the neuroanatomical data. To tie the progression in brain atrophy to a behavioral marker, we analyzed the ADAS-cog sub-scores in the stages. RESULTS: The optimal model consists of eight states with differentiable neuroanatomical features, forming two routes crossing once at a very early point and merging at the final state. The cortical route is characterized by early and sustained atrophy in cortical regions. The limbic route is characterized by early decrease in limbic regions. Cognitive differences between the two routes are most noticeable in the memory domain with subjects from the limbic route experiencing stronger memory impairments. CONCLUSION: Our findings corroborate that more than one pattern of grey matter deterioration with several discernable stages can be identified in the progression of AD. These neuroanatomical subtypes are behaviorally meaningful and provide a door into early diagnosis of AD and prognosis of the disease's progression.

Factors governing the assignment of visual consequence to the corresponding action.

To achieve fast feedback control of voluntary movements, the visual consequences of our motor commands need to be quickly identified and analyzed by the motor control processes in the brain. These processes work remarkably well even in complex visual environments and in the face of discrepancies between physical actuator and visually perceived effect, e.g. when moving a computer mouse on a visually crowded screen. Here, we use an ambiguous situation in which a single cursor could be controlled by either the left or the right hand to determine the visual and cognitive factors that determine the assignment of a visual stimulus to the corresponding motor command. Our results demonstrate that the visuomotor system is exquisitely sensitive to the spatio-temporal correlation between cursor and hands, learning the appropriate mapping implicitly within several minutes. In contrast, spatial proximity between end effector and visual consequence has an immediate but only transient effect on the assignment process. Finally, an explicit instruction about which hand controls the cursor only has a minor influence when the instruction is presented first. These findings provide insight into the relative importance of the factors that determine the binding of visual information to the corresponding motor structures to enable fast feedback control.NEW & NOTEWORTHY For efficient visuomotor online control, the brain needs to solve the correspondence problem between an ongoing movement and its visual consequences. Here, we challenge the visuomotor system with an ambiguous reaching task, in which the visual feedback was controlled by either hand or by a combination of both. Our findings characterize the properties of a flexible assignment process that quickly takes into account the spatio-temporal properties of movements and the visual scene.

Minimizing endpoint variability through reinforcement learning during reaching movements involving shoulder, elbow and wrist.

Reaching movements are comprised of the coordinated action across multiple joints. The human skeleton is redundant for this task because different joint configurations can lead to the same endpoint in space. How do people learn to use combinations of joints that maximize success in goal-directed motor tasks? To answer this question, we used a 3-degree-of-freedom manipulandum to measure shoulder, elbow and wrist joint movements during reaching in a plane. We tested whether a shift in the relative contribution of the wrist and elbow joints to a reaching movement could be learned by an implicit reinforcement regime. Unknown to the participants, we decreased the task success for certain joint configurations (wrist flexion or extension, respectively) by adding random variability to the endpoint feedback. In return, the opposite wrist postures were rewarded in the two experimental groups (flexion and extension group). We found that the joint configuration slowly shifted towards movements that provided more control over the endpoint and hence higher task success. While the overall learning was significant, only the group that was guided to extend the wrist joint more during the movement showed substantial learning. Importantly, all changes in movement pattern occurred independent of conscious awareness of the experimental manipulation. These findings suggest that the motor system is generally sensitive to its output variability and can optimize joint-space solutions that minimize task-relevant output variability. We discuss biomechanical biases (e.g. joint's range of movement) that could impose hurdles to the learning process.

Temporal Evolution of Spatial Computations for Visuomotor Control.

Goal-directed reaching movements are guided by visual feedback from both target and hand. The classical view is that the brain extracts information about target and hand positions from a visual scene, calculates a difference vector between them, and uses this estimate to control the movement. Here we show that during fast feedback control, this computation is not immediate, but evolves dynamically over time. Immediately after a change in the visual scene, the motor system generates independent responses to the errors in hand and target location. Only about 200 ms later, the changes in target and hand positions are combined appropriately in the response, slowly converging to the true difference vector. Therefore, our results provide evidence for the temporal evolution of spatial computations in the human visuomotor system, in which the accurate difference vector computation is first estimated by a fast approximation.

Reaching with the sixth sense: Vestibular contributions to voluntary motor control in the human right parietal cortex.

The vestibular system constitutes the silent sixth sense: It automatically triggers a variety of vital reflexes to maintain postural and visual stability. Beyond their role in reflexive behavior, vestibular afferents contribute to several perceptual and cognitive functions and also support voluntary control of movements by complementing the other senses to accomplish the movement goal. Investigations into the neural correlates of vestibular contribution to voluntary action in humans are challenging and have progressed far less than research on corresponding visual and proprioceptive involvement. Here, we demonstrate for the first time with event-related TMS that the posterior part of the right medial intraparietal sulcus processes vestibular signals during a goal-directed reaching task with the dominant right hand. This finding suggests a qualitative difference between the processing of vestibular vs. visual and proprioceptive signals for controlling voluntary movements, which are pre-dominantly processed in the left posterior parietal cortex. Furthermore, this study reveals a neural pathway for vestibular input that might be distinct from the processing for reflexive or cognitive functions, and opens a window into their investigation in humans.

Processing reafferent and exafferent visual information for action and perception.

A recent study suggests that reafferent hand-related visual information utilizes a privileged, attention-independent processing channel for motor control. This process was termed visuomotor binding to reflect its proposed function: linking visual reafferences to the corresponding motor control centers. Here, we ask whether the advantage of processing reafferent over exafferent visual information is a specific feature of the motor processing stream or whether the improved processing also benefits the perceptual processing stream. Human participants performed a bimanual reaching task in a cluttered visual display, and one of the visual hand cursors could be displaced laterally during the movement. We measured the rapid feedback responses of the motor system as well as matched perceptual judgments of which cursor was displaced. Perceptual judgments were either made by watching the visual scene without moving or made simultaneously to the reaching tasks, such that the perceptual processing stream could also profit from the specialized processing of reafferent information in the latter case. Our results demonstrate that perceptual judgments in the heavily cluttered visual environment were improved when performed based on reafferent information. Even in this case, however, the filtering capability of the perceptual processing stream suffered more from the increasing complexity of the visual scene than the motor processing stream. These findings suggest partly shared and partly segregated processing of reafferent information for vision for motor control versus vision for perception.

A dedicated binding mechanism for the visual control of movement.

The human motor system is remarkably proficient in the online control of visually guided movements, adjusting to changes in the visual scene within 100 ms [1-3]. This is achieved through a set of highly automatic processes [4] translating visual information into representations suitable for motor control [5, 6]. For this to be accomplished, visual information pertaining to target and hand need to be identified and linked to the appropriate internal representations during the movement. Meanwhile, other visual information must be filtered out, which is especially demanding in visually cluttered natural environments. If selection of relevant sensory information for online control was achieved by visual attention, its limited capacity [7] would substantially constrain the efficiency of visuomotor feedback control. Here we demonstrate that both exogenously and endogenously cued attention facilitate the processing of visual target information [8], but not of visual hand information. Moreover, distracting visual information is more efficiently filtered out during the extraction of hand compared to target information. Our results therefore suggest the existence of a dedicated visuomotor binding mechanism that links the hand representation in visual and motor systems.

A key region in the human parietal cortex for processing proprioceptive hand feedback during reaching movements.

Seemingly effortless, we adjust our movements to continuously changing environments. After initiation of a goal-directed movement, the motor command is under constant control of sensory feedback loops. The main sensory signals contributing to movement control are vision and proprioception. Recent neuroimaging studies have focused mainly on identifying the parts of the posterior parietal cortex (PPC) that contribute to visually guided movements. We used event-related TMS and force perturbations of the reaching hand to test whether the same sub-regions of the left PPC contribute to the processing of proprioceptive-only and of multi-sensory information about hand position when reaching for a visual target. TMS over two distinct stimulation sites elicited differential effects: TMS applied over the posterior part of the medial intraparietal sulcus (mIPS) compromised reaching accuracy when proprioception was the only sensory information available for correcting the reaching error. When visual feedback of the hand was available, TMS over the anterior intraparietal sulcus (aIPS) prolonged reaching time. Our results show for the first time the causal involvement of the posterior mIPS in processing proprioceptive feedback for online reaching control, and demonstrate that distinct cortical areas process proprioceptive-only and multi-sensory information for fast feedback corrections.

Mechanisms of responsibility assignment during redundant reaching movements.

When the two hands act together to achieve a goal, the redundancy of the system makes it necessary to distribute the responsibility for error corrections across the two hands. In an experiment in which participants control a single cursor with the movements of both hands, we show that right-handed individuals correct for movement errors more with their nondominant left hand than with their right hand, even though the dominant right hand corrects the same errors more quickly and efficiently when each hand acts in isolation. By measuring the responses to rapid cursor and target displacements using force channels, we demonstrate that this shift is due to a modulation of the feedback gains of each hand rather than to a shift in the onset of the corrective response. We also show that the shift toward left-hand corrections is more pronounced for errors that lead to adaptation (cursor displacements) than for perturbations that do not (target displacements). This finding provides some support for the idea that the motor system assigns the correction to the most likely source of the error to induce learning and to optimize future performance. Finally, we find that the relative strength of the feedback corrections in the redundant task correlates positively with those found for the nonredundant tasks. Thus the process of responsibility assignment modulates the processes that normally determine the gains of feedback correction rather than completely overwriting them.

Contributions of the PPC to online control of visually guided reaching movements assessed with fMRI-guided TMS.

The posterior parietal cortex (PPC) plays an important role in controlling voluntary movements by continuously integrating sensory information about body state and the environment. We tested which subregions of the PPC contribute to the processing of target- and body-related visual information while reaching for an object, using a reaching paradigm with 2 types of visual perturbation: displacement of the visual target and displacement of the visual feedback about the hand position. Initially, functional magnetic resonance imaging (fMRI) was used to localize putative target areas involved in online corrections of movements in response to perturbations. The causal contribution of these areas to online correction was tested in subsequent neuronavigated transcranial magnetic stimulation (TMS) experiments. Robust TMS effects occurred at distinct anatomical sites along the anterior intraparietal sulcus (aIPS) and the anterior part of the supramarginal gyrus for both perturbations. TMS over neighboring sites did not affect online control. Our results support the hypothesis that the aIPS is more generally involved in visually guided control of movements, independent of body effectors and nature of the visual information. Furthermore, they suggest that the human network of PPC subregions controlling goal-directed visuomotor processes extends more inferiorly than previously thought. Our results also point toward a good spatial specificity of the TMS effects.

Seeing the hand while reaching speeds up on-line responses to a sudden change in target position.

Goal-directed movements are executed under the permanent supervision of the central nervous system, which continuously processes sensory afferents and triggers on-line corrections if movement accuracy seems to be compromised. For arm reaching movements, visual information about the hand plays an important role in this supervision, notably improving reaching accuracy. Here, we tested whether visual feedback of the hand affects the latency of on-line responses to an external perturbation when reaching for a visual target. Two types of perturbation were used: visual perturbation consisted in changing the spatial location of the target and kinesthetic perturbation in applying a force step to the reaching arm. For both types of perturbation, the hand trajectory and the electromyographic (EMG) activity of shoulder muscles were analysed to assess whether visual feedback of the hand speeds up on-line corrections. Without visual feedback of the hand, on-line responses to visual perturbation exhibited the longest latency. This latency was reduced by about 10% when visual feedback of the hand was provided. On the other hand, the latency of on-line responses to kinesthetic perturbation was independent of the availability of visual feedback of the hand. In a control experiment, we tested the effect of visual feedback of the hand on visual and kinesthetic two-choice reaction times--for which coordinate transformation is not critical. Two-choice reaction times were never facilitated by visual feedback of the hand. Taken together, our results suggest that visual feedback of the hand speeds up on-line corrections when the position of the visual target with respect to the body must be re-computed during movement execution. This facilitation probably results from the possibility to map hand- and target-related information in a common visual reference frame.