A Machine Learning-Based Holistic Approach to Predict the Clinical Course of Patients within the Alzheimer's Disease Spectrum.

BACKGROUND: Alzheimer's disease (AD) is a neurodegenerative condition driven by multifactorial etiology. Mild cognitive impairment (MCI) is a transitional condition between healthy aging and dementia. No reliable biomarkers are available to predict the conversion from MCI to AD. OBJECTIVE: To evaluate the use of machine learning (ML) on a wealth of data offered by the Alzheimer's Disease Neuroimaging Initiative (ADNI) and Alzheimer's Disease Metabolomics Consortium (ADMC) database in the prediction of the MCI to AD conversion. METHODS: We implemented an ML-based Random Forest (RF) algorithm to predict conversion from MCI to AD. Data related to the study population (587 MCI subjects) were analyzed by RF as separate or combined features and assessed for classification power. Four classes of variables were considered: neuropsychological test scores, AD-related cerebrospinal fluid (CSF) biomarkers, peripheral biomarkers, and structural magnetic resonance imaging (MRI) variables. RESULTS: The ML-based algorithm exhibited 86% accuracy in predicting the AD conversion of MCI subjects. When assessing the features that helped the most, neuropsychological test scores, MRI data, and CSF biomarkers were the most relevant in the MCI to AD prediction. Peripheral parameters were effective when employed in association with neuropsychological test scores. Age and sex differences modulated the prediction accuracy. AD conversion was more effectively predicted in females and younger subjects. CONCLUSION: Our findings support the notion that AD-related neurodegenerative processes result from the concerted activity of multiple pathological mechanisms and factors that act inside and outside the brain and are dynamically affected by age and sex.

ECoG Beta Suppression and Modulation During Finger Extension and Flexion.

Neural oscillations originate predominantly from interacting cortical neurons and consequently reflect aspects of cortical information processing. However, their functional role is not yet fully understood and their interpretation is debatable. Amplitude modulations (AMs) in alpha (8-12 Hz), beta (13-30 Hz), and high gamma (70-150 Hz) band in invasive electrocorticogram (ECoG) and non-invasive electroencephalogram (EEG) signals change with behavior. Alpha and beta band AMs are typically suppressed (desynchronized) during motor behavior, while high gamma AMs highly correlate with the behavior. These two phenomena are successfully used for functional brain mapping and brain-computer interface (BCI) applications. Recent research found movement-phase related AMs (MPA) also in high beta/low gamma (24-40 Hz) EEG rhythms. These MPAs were found by separating the suppressed AMs into sustained and dynamic components. Sustained AM components are those with frequencies that are lower than the motor behavior. Dynamic components those with frequencies higher than the behavior. In this paper, we study ECoG beta/low gamma band (12-30 Hz/30-42 Hz) AM during repetitive finger movements addressing the question whether or not MPAs can be found in ECoG beta band. Indeed, MPA in the 12-18 Hz and 18-24 Hz band were found. This additional information may lead to further improvements in ECoG-based prediction and reconstruction of motor behavior by combining high gamma AM and beta band MPA.

Thought-based interaction: Same data, same methods, different results?

Restoration of communication in people with complete motor paralysis-a condition called complete locked-in state (CLIS)-is one of the greatest challenges of brain-computer interface (BCI) research. New findings have recently been presented that bring us one step closer to this goal. However, the validity of the evidence has been questioned: independent reanalysis of the same data yielded significantly different results. Reasons for the failure to replicate the findings must be of a methodological nature. What is the best practice to ensure that results are stringent and conclusive and analyses replicable? Confirmation bias and the counterintuitive nature of probability may lead to an overly optimistic interpretation of new evidence. Lack of detail complicates replicability.

Workshops of the Sixth International Brain-Computer Interface Meeting: brain-computer interfaces past, present, and future.

The Sixth International Brain-Computer Interface (BCI) Meeting was held 30 May-3 June 2016 at the Asilomar Conference Grounds, Pacific Grove, California, USA. The conference included 28 workshops covering topics in BCI and brain-machine interface research. Topics included BCI for specific populations or applications, advancing BCI research through use of specific signals or technological advances, and translational and commercial issues to bring both implanted and non-invasive BCIs to market. BCI research is growing and expanding in the breadth of its applications, the depth of knowledge it can produce, and the practical benefit it can provide both for those with physical impairments and the general public. Here we provide summaries of each workshop, illustrating the breadth and depth of BCI research and highlighting important issues and calls for action to support future research and development.

Heading for new shores! Overcoming pitfalls in BCI design.

Research in brain-computer interfaces has achieved impressive progress towards implementing assistive technologies for restoration or substitution of lost motor capabilities, as well as supporting technologies for able-bodied subjects. Notwithstanding this progress, effective translation of these interfaces from proof-of concept prototypes into reliable applications remains elusive. As a matter of fact, most of the current BCI systems cannot be used independently for long periods of time by their intended end-users. Multiple factors that impair achieving this goal have already been identified. However, it is not clear how do they affect the overall BCI performance or how they should be tackled. This is worsened by the publication bias where only positive results are disseminated, preventing the research community from learning from its errors. This paper is the result of a workshop held at the 6th International BCI meeting in Asilomar. We summarize here the discussion on concrete research avenues and guidelines that may help overcoming common pitfalls and make BCIs become a useful alternative communication device.

EEG Oscillations Are Modulated in Different Behavior-Related Networks during Rhythmic Finger Movements.

Sequencing and timing of body movements are essential to perform motoric tasks. In this study, we investigate the temporal relation between cortical oscillations and human motor behavior (i.e., rhythmic finger movements). High-density EEG recordings were used for source imaging based on individual anatomy. We separated sustained and movement phase-related EEG source amplitudes based on the actual finger movements recorded by a data glove. Sustained amplitude modulations in the contralateral hand area show decrease for α (10-12 Hz) and ß (18-24 Hz), but increase for high γ (60-80 Hz) frequencies during the entire movement period. Additionally, we found movement phase-related amplitudes, which resembled the flexion and extension sequence of the fingers. Especially for faster movement cadences, movement phase-related amplitudes included high ß (24-30 Hz) frequencies in prefrontal areas. Interestingly, the spectral profiles and source patterns of movement phase-related amplitudes differed from sustained activities, suggesting that they represent different frequency-specific large-scale networks. First, networks were signified by the sustained element, which statically modulate their synchrony levels during continuous movements. These networks may upregulate neuronal excitability in brain regions specific to the limb, in this study the right hand area. Second, movement phase-related networks, which modulate their synchrony in relation to the movement sequence. We suggest that these frequency-specific networks are associated with distinct functions, including top-down control, sensorimotor prediction, and integration. The separation of different large-scale networks, we applied in this work, improves the interpretation of EEG sources in relation to human motor behavior. SIGNIFICANCE STATEMENT: EEG recordings provide high temporal resolution suitable to relate cortical oscillations to actual movements. Investigating EEG sources during rhythmic finger movements, we distinguish sustained from movement phase-related amplitude modulations. We separate these two EEG source elements motivated by our previous findings in gait. Here, we found two types of large-scale networks, representing the right fingers in distinction from the time sequence of the movements. These findings suggest that EEG source amplitudes reconstructed in a cortical patch are the superposition of these simultaneously present network activities. Separating these frequency-specific networks is relevant for studying function and possible dysfunction of the cortical sensorimotor system in humans as well as to provide more advanced features for brain-computer interfaces.

Random forests in non-invasive sensorimotor rhythm brain-computer interfaces: a practical and convenient non-linear classifier.

There is general agreement in the brain-computer interface (BCI) community that although non-linear classifiers can provide better results in some cases, linear classifiers are preferable. Particularly, as non-linear classifiers often involve a number of parameters that must be carefully chosen. However, new non-linear classifiers were developed over the last decade. One of them is the random forest (RF) classifier. Although popular in other fields of science, RFs are not common in BCI research. In this work, we address three open questions regarding RFs in sensorimotor rhythm (SMR) BCIs: parametrization, online applicability, and performance compared to regularized linear discriminant analysis (LDA). We found that the performance of RF is constant over a large range of parameter values. We demonstrate - for the first time - that RFs are applicable online in SMR-BCIs. Further, we show in an offline BCI simulation that RFs statistically significantly outperform regularized LDA by about 3%. These results confirm that RFs are practical and convenient non-linear classifiers for SMR-BCIs. Taking into account further properties of RFs, such as independence from feature distributions, maximum margin behavior, multiclass and advanced data mining capabilities, we argue that RFs should be taken into consideration for future BCIs.

Corrigendum: EEG beta suppression and low gamma modulation are different elements of human upright walking.

[This corrects the article on p. 485 in vol. 8, PMID: 25071515.].

Individually adapted imagery improves brain-computer interface performance in end-users with disability.

Brain-computer interfaces (BCIs) translate oscillatory electroencephalogram (EEG) patterns into action. Different mental activities modulate spontaneous EEG rhythms in various ways. Non-stationarity and inherent variability of EEG signals, however, make reliable recognition of modulated EEG patterns challenging. Able-bodied individuals who use a BCI for the first time achieve - on average - binary classification performance of about 75%. Performance in users with central nervous system (CNS) tissue damage is typically lower. User training generally enhances reliability of EEG pattern generation and thus also robustness of pattern recognition. In this study, we investigated the impact of mental tasks on binary classification performance in BCI users with central nervous system (CNS) tissue damage such as persons with stroke or spinal cord injury (SCI). Motor imagery (MI), that is the kinesthetic imagination of movement (e.g. squeezing a rubber ball with the right hand), is the "gold standard" and mainly used to modulate EEG patterns. Based on our recent results in able-bodied users, we hypothesized that pair-wise combination of "brain-teaser" (e.g. mental subtraction and mental word association) and "dynamic imagery" (e.g. hand and feet MI) tasks significantly increases classification performance of induced EEG patterns in the selected end-user group. Within-day (How stable is the classification within a day?) and between-day (How well does a model trained on day one perform on unseen data of day two?) analysis of variability of mental task pair classification in nine individuals confirmed the hypothesis. We found that the use of the classical MI task pair hand vs. feed leads to significantly lower classification accuracy - in average up to 15% less - in most users with stroke or SCI. User-specific selection of task pairs was again essential to enhance performance. We expect that the gained evidence will significantly contribute to make imagery-based BCI technology become accessible to a larger population of users including individuals with special needs due to CNS damage.

High and low gamma EEG oscillations in central sensorimotor areas are conversely modulated during the human gait cycle.

Investigating human brain function is essential to develop models of cortical involvement during walking. Such models could advance the analysis of motor impairments following brain injuries (e.g., stroke) and may lead to novel rehabilitation approaches. In this work, we applied high-density EEG source imaging based on individual anatomy to enable neuroimaging during walking. To minimize the impact of muscular influence on EEG recordings we introduce a novel artifact correction method based on spectral decomposition. High γ oscillations (>60Hz) were previously reported to play an important role in motor control. Here, we investigate high γ amplitudes while focusing on two different aspects of a walking experiment, namely the fact that a person walks and the rhythmicity of walking. We found that high γ amplitudes (60-80Hz), located focally in central sensorimotor areas, were significantly increased during walking compared to standing. Moreover, high γ (70-90Hz) amplitudes in the same areas are modulated in relation to the gait cycle. Since the spectral peaks of high γ amplitude increase and modulation do not match, it is plausible that these two high γ elements represent different frequency-specific network interactions. Interestingly, we found high γ (70-90Hz) amplitudes to be coupled to low γ (24-40Hz) amplitudes, which both are modulated in relation to the gait cycle but conversely to each other. In summary, our work is a further step towards modeling cortical involvement during human upright walking.

Thought-based row-column scanning communication board for individuals with cerebral palsy.

Impairment of an individual's ability to communicate is a major hurdle for active participation in education and social life. A lot of individuals with cerebral palsy (CP) have normal intelligence, however, due to their inability to communicate, they fall behind. Non-invasive electroencephalogram (EEG) based brain-computer interfaces (BCIs) have been proposed as potential assistive devices for individuals with CP. BCIs translate brain signals directly into action. Motor activity is no longer required. However, translation of EEG signals may be unreliable and requires months of training. Moreover, individuals with CP may exhibit high levels of spontaneous and uncontrolled movement, which has a large impact on EEG signal quality and results in incorrect translations. We introduce a novel thought-based row-column scanning communication board that was developed following user-centered design principles. Key features include an automatic online artifact reduction method and an evidence accumulation procedure for decision making. The latter allows robust decision making with unreliable BCI input. Fourteen users with CP participated in a supporting online study and helped to evaluate the performance of the developed system. Users were asked to select target items with the row-column scanning communication board. The results suggest that seven among eleven remaining users performed better than chance and were consequently able to communicate by using the developed system. Three users were excluded because of insufficient EEG signal quality. These results are very encouraging and represent a good foundation for the development of real-world BCI-based communication devices for users with CP.

Bring mental activity into action! An enhanced online co-adaptive brain-computer interface training protocol.

Non-stationarity and inherent variability of the noninvasive electroencephalogram (EEG) makes robust recognition of spontaneous EEG patterns challenging. Reliable modulation of EEG patterns that a BCI can robustly detect is a skill that users must learn. In this paper, we present a novel online co-adaptive BCI training paradigm. The system autonomously screens users for their ability to modulate EEG patterns in a predictive way and adapts its model parameters online. Results of a supporting study in seven first-time BCI users with disability are very encouraging. Three of 7 users achieved online accuracy > 70% for 2-class BCI control after 24 minutes of training. Online performance in 6 of 7 users was significantly higher than chance level. Online control was based on one single bipolar EEG channel. Beta band activity carried most discriminant information. Our fully automatic co-adaptive online approach allows to evaluate whether user benefit from current BCI technology within a reasonable timescale.

FORCe: Fully Online and Automated Artifact Removal for Brain-Computer Interfacing.

A fully automated and online artifact removal method for the electroencephalogram (EEG) is developed for use in brain-computer interfacing (BCI). The method (FORCe) is based upon a novel combination of wavelet decomposition, independent component analysis, and thresholding. FORCe is able to operate on a small channel set during online EEG acquisition and does not require additional signals (e.g., electrooculogram signals). Evaluation of FORCe is performed offline on EEG recorded from 13 BCI particpants with cerebral palsy (CP) and online with three healthy participants. The method outperforms the state-of the-art automated artifact removal methods Lagged Auto-Mutual Information Clustering (LAMIC) and Fully Automated Statistical Thresholding for EEG artifact Rejection (FASTER), and is able to remove a wide range of artifact types including blink, electromyogram (EMG), and electrooculogram (EOG) artifacts.

A co-adaptive sensory motor rhythms Brain-Computer Interface based on common spatial patterns and Random Forest.

Sensorimotor rhythm (SMR) based Brain-Computer Interfaces (BCI) typically require lengthy user training. This can be exhausting and fatiguing for the user as data collection may be monotonous and typically without any feedback for user motivation. Hence new ways to reduce user training and improve performance are needed. We recently introduced a two class motor imagery BCI system which continuously adapted with increasing run-time to the brain patterns of the user. The system was designed to provide visual feedback to the user after just five minutes. The aim of the current work was to improve user-specific online adaptation, which was expected to lead to higher performances. To maximize SMR discrimination, the method of filter-bank common spatial patterns (fbCSP) and Random Forest (RF) classifier were combined. In a supporting online study, all volunteers performed significantly better than chance. Overall peak accuracy of 88.6 ± 6.1 (SD) % was reached, which significantly exceeded the performance of our previous system by 13%. Therefore, we consider this system the next step towards fully auto-calibrating motor imagery BCIs.

Grip force modulation characteristics as a marker for clinical disease progression in individuals with Parkinson disease: case-control study.

BACKGROUND: Upper extremity deficits are prevalent in individuals with Parkinson disease (PD). In the early stages of PD, such deficits can be subtle and challenging to document on clinical examination. OBJECTIVE: The purpose of this study was to use a novel force sensor system to characterize grip force modulation, including force, temporal, and movement quality parameters, during a fine motor control task in individuals with early stage PD. DESIGN: A case-control study was conducted. METHODS: Fourteen individuals with early stage PD were compared with a control group of 14 healthy older adults. The relationship of force modulation parameters with motor symptom severity and disease chronicity also was assessed in people with PD. Force was measured during both precision and power grasp tasks using an instrumented twist-cap device capable of rotating in either direction. RESULTS: Compared with the control group, the PD group demonstrated more movement arrests during both precision and power grasp and longer total movement times during the power grasp. These deficits persisted when a concurrent cognitive task was added, with some evidence of force control deficits in the PD group, including lower rates of force production during the precision grasp task and higher peak forces during the power grasp task. For precision grasp, a higher number of movement arrests in single- and dual-task conditions as well as longer total movement times in the dual-task condition were associated with more severe motor symptoms. LIMITATIONS: The sample was small and consisted of individuals in the early stages of PD with mild motor deficits. The group with PD was predominantly male, whereas the control group was predominantly female. CONCLUSION: The results suggest that assessing grip force modulation deficits during fine motor tasks is possible with instrumented devices, and such sensitive measures may be important for detecting and tracking change early in the progression of PD.

Non-motor tasks improve adaptive brain-computer interface performance in users with severe motor impairment.

Individuals with severe motor impairment can use event-related desynchronization (ERD) based BCIs as assistive technology. Auto-calibrating and adaptive ERD-based BCIs that users control with motor imagery tasks ("SMR-AdBCI") have proven effective for healthy users. We aim to find an improved configuration of such an adaptive ERD-based BCI for individuals with severe motor impairment as a result of spinal cord injury (SCI) or stroke. We hypothesized that an adaptive ERD-based BCI, that automatically selects a user specific class-combination from motor-related and non motor-related mental tasks during initial auto-calibration ("Auto-AdBCI") could allow for higher control performance than a conventional SMR-AdBCI. To answer this question we performed offline analyses on two sessions (21 data sets total) of cue-guided, five-class electroencephalography (EEG) data recorded from individuals with SCI or stroke. On data from the twelve individuals in Session 1, we first identified three bipolar derivations for the SMR-AdBCI. In a similar way, we determined three bipolar derivations and four mental tasks for the Auto-AdBCI. We then simulated both, the SMR-AdBCI and the Auto-AdBCI configuration on the unseen data from the nine participants in Session 2 and compared the results. On the unseen data of Session 2 from individuals with SCI or stroke, we found that automatically selecting a user specific class-combination from motor-related and non motor-related mental tasks during initial auto-calibration (Auto-AdBCI) significantly (p < 0.01) improved classification performance compared to an adaptive ERD-based BCI that only used motor imagery tasks (SMR-AdBCI; average accuracy of 75.7 vs. 66.3%).

A co-adaptive brain-computer interface for end users with severe motor impairment.

Co-adaptive training paradigms for event-related desynchronization (ERD) based brain-computer interfaces (BCI) have proven effective for healthy users. As of yet, it is not clear whether co-adaptive training paradigms can also benefit users with severe motor impairment. The primary goal of our paper was to evaluate a novel cue-guided, co-adaptive BCI training paradigm with severely impaired volunteers. The co-adaptive BCI supports a non-control state, which is an important step toward intuitive, self-paced control. A secondary aim was to have the same participants operate a specifically designed self-paced BCI training paradigm based on the auto-calibrated classifier. The co-adaptive BCI analyzed the electroencephalogram from three bipolar derivations (C3, Cz, and C4) online, while the 22 end users alternately performed right hand movement imagery (MI), left hand MI and relax with eyes open (non-control state). After less than five minutes, the BCI auto-calibrated and proceeded to provide visual feedback for the MI task that could be classified better against the non-control state. The BCI continued to regularly recalibrate. In every calibration step, the system performed trial-based outlier rejection and trained a linear discriminant analysis classifier based on one auto-selected logarithmic band-power feature. In 24 minutes of training, the co-adaptive BCI worked significantly (p = 0.01) better than chance for 18 of 22 end users. The self-paced BCI training paradigm worked significantly (p = 0.01) better than chance in 11 of 20 end users. The presented co-adaptive BCI complements existing approaches in that it supports a non-control state, requires very little setup time, requires no BCI expert and works online based on only two electrodes. The preliminary results from the self-paced BCI paradigm compare favorably to previous studies and the collected data will allow to further improve self-paced BCI systems for disabled users.

EEG beta suppression and low gamma modulation are different elements of human upright walking.

Cortical involvement during upright walking is not well-studied in humans. We analyzed non-invasive electroencephalographic (EEG) recordings from able-bodied volunteers who participated in a robot-assisted gait-training experiment. To enable functional neuroimaging during walking, we applied source modeling to high-density (120 channels) EEG recordings using individual anatomy reconstructed from structural magnetic resonance imaging scans. First, we analyzed amplitude differences between the conditions, walking and upright standing. Second, we investigated amplitude modulations related to the gait phase. During active walking upper µ (10-12 Hz) and ß (18-30 Hz) oscillations were suppressed [event-related desynchronization (ERD)] compared to upright standing. Significant ß ERD activity was located focally in central sensorimotor areas for 9/10 subjects. Additionally, we found that low γ (24-40 Hz) amplitudes were modulated related to the gait phase. Because there is a certain frequency band overlap between sustained ß ERD and gait phase related modulations in the low γ range, these two phenomena are superimposed. Thus, we observe gait phase related amplitude modulations at a certain ERD level. We conclude that sustained µ and ß ERD reflect a movement related state change of cortical excitability while gait phase related modulations in the low γ represent the motion sequence timing during gait. Interestingly, the center frequencies of sustained ß ERD and gait phase modulated amplitudes were identified to be different. They may therefore be caused by different neuronal rhythms, which should be taken under consideration in future studies.

Exploration of the neural correlates of cerebral palsy for sensorimotor BCI control.

Cerebral palsy (CP) includes a broad range of disorders, which can result in impairment of posture and movement control. Brain-computer interfaces (BCIs) have been proposed as assistive devices for individuals with CP. Better understanding of the neural processing underlying motor control in affected individuals could lead to more targeted BCI rehabilitation and treatment options. We have explored well-known neural correlates of movement, including event-related desynchronization (ERD), phase synchrony, and a recently-introduced measure of phase dynamics, in participants with CP and healthy control participants. Although present, significantly less ERD and phase locking were found in the group with CP. Additionally, inter-group differences in phase dynamics were also significant. Taken together these findings suggest that users with CP exhibit lower levels of motor cortex activation during motor imagery, as reflected in lower levels of ongoing mu suppression and less functional connectivity. These differences indicate that development of BCIs for individuals with CP may pose additional challenges beyond those faced in providing BCIs to healthy individuals.