Impact of dataset size and long-term ECoG-based BCI usage on deep learning decoders performance.

Introduction: In brain-computer interfaces (BCI) research, recording data is time-consuming and expensive, which limits access to big datasets. This may influence the BCI system performance as machine learning methods depend strongly on the training dataset size. Important questions arise: taking into account neuronal signal characteristics (e.g., non-stationarity), can we achieve higher decoding performance with more data to train decoders? What is the perspective for further improvement with time in the case of long-term BCI studies? In this study, we investigated the impact of long-term recordings on motor imagery decoding from two main perspectives: model requirements regarding dataset size and potential for patient adaptation. Methods: We evaluated the multilinear model and two deep learning (DL) models on a long-term BCI & Tetraplegia (ClinicalTrials.gov identifier: NCT02550522) clinical trial dataset containing 43 sessions of ECoG recordings performed with a tetraplegic patient. In the experiment, a participant executed 3D virtual hand translation using motor imagery patterns. We designed multiple computational experiments in which training datasets were increased or translated to investigate the relationship between models' performance and different factors influencing recordings. Results: Our results showed that DL decoders showed similar requirements regarding the dataset size compared to the multilinear model while demonstrating higher decoding performance. Moreover, high decoding performance was obtained with relatively small datasets recorded later in the experiment, suggesting motor imagery patterns improvement and patient adaptation during the long-term experiment. Finally, we proposed UMAP embeddings and local intrinsic dimensionality as a way to visualize the data and potentially evaluate data quality. Discussion: DL-based decoding is a prospective approach in BCI which may be efficiently applied with real-life dataset size. Patient-decoder co-adaptation is an important factor to consider in long-term clinical BCI.

Decoding ECoG signal into 3D hand translation using deep learning.

Objective.Motor brain-computer interfaces (BCIs) are a promising technology that may enable motor-impaired people to interact with their environment. BCIs would potentially compensate for arm and hand function loss, which is the top priority for individuals with tetraplegia. Designing real-time and accurate BCI is crucial to make such devices useful, safe, and easy to use by patients in a real-life environment. Electrocorticography (ECoG)-based BCIs emerge as a good compromise between invasiveness of the recording device and good spatial and temporal resolution of the recorded signal. However, most ECoG signal decoders used to predict continuous hand movements are linear models. These models have a limited representational capacity and may fail to capture the relationship between ECoG signal features and continuous hand movements. Deep learning (DL) models, which are state-of-the-art in many problems, could be a solution to better capture this relationship.Approach.In this study, we tested several DL-based architectures to predict imagined 3D continuous hand translation using time-frequency features extracted from ECoG signals. The dataset used in the analysis is a part of a long-term clinical trial (ClinicalTrials.gov identifier: NCT02550522) and was acquired during a closed-loop experiment with a tetraplegic subject. The proposed architectures include multilayer perceptron, convolutional neural networks (CNNs), and long short-term memory networks (LSTM). The accuracy of the DL-based and multilinear models was compared offline using cosine similarity.Main results.Our results show that CNN-based architectures outperform the current state-of-the-art multilinear model. The best architecture exploited the spatial correlation between neighboring electrodes with CNN and benefited from the sequential character of the desired hand trajectory by using LSTMs. Overall, DL increased the average cosine similarity, compared to the multilinear model, by up to 60%, from 0.189 to 0.302 and from 0.157 to 0.249 for the left and right hand, respectively.Significance.This study shows that DL-based models could increase the accuracy of BCI systems in the case of 3D hand translation prediction in a tetraplegic subject.

An adaptive closed-loop ECoG decoder for long-term and stable bimanual control of an exoskeleton by a tetraplegic.

Objective.The article aims at addressing 2 challenges to step motor brain-computer interface (BCI) out of laboratories: asynchronous control of complex bimanual effectors with large numbers of degrees of freedom, using chronic and safe recorders, and the decoding performance stability over time without frequent decoder recalibration.Approach.Closed-loop adaptive/incremental decoder training is one strategy to create a model stable over time. Adaptive decoders update their parameters with new incoming data, optimizing the model parameters in real time. It allows cross-session training with multiple recording conditions during closed loop BCI experiments. In the article, an adaptive tensor-based recursive exponentially weighted Markov-switching multi-linear model (REW-MSLM) decoder is proposed. REW-MSLM uses a mixture of expert (ME) architecture, mixing or switching independent decoders (experts) according to the probability estimated by a 'gating' model. A Hidden Markov model approach is employed as gating model to improve the decoding robustness and to provide strong idle state support. The ME architecture fits the multi-limb paradigm associating an expert to a particular limb or action.Main results.Asynchronous control of an exoskeleton by a tetraplegic patient using a chronically implanted epidural electrocorticography (EpiCoG) recorder is reported. The stable over a period of six months (without decoder recalibration) eight-dimensional alternative bimanual control of the exoskeleton and its virtual avatar is demonstrated.Significance.Based on the long-term (>36 months) chronic bilateral EpiCoG recordings in a tetraplegic (ClinicalTrials.gov, NCT02550522), we addressed the poorly explored field of asynchronous bimanual BCI. The new decoder was designed to meet to several challenges: the high-dimensional control of a complex effector in experiments closer to real-world behavior (point-to-point pursuit versus conventional center-out tasks), with the ability of the BCI system to act as a stand-alone device switching between idle and control states, and a stable performance over a long period of time without decoder recalibration.

Automatic segmentation and location learning of neonatal cerebral ventricles in 3D ultrasound data combining CNN and CPPN.

Preterm neonates are highly likely to suffer from ventriculomegaly, a dilation of the Cerebral Ventricular System (CVS). This condition can develop into life-threatening hydrocephalus and is correlated with future neuro-developmental impairments. Consequently, it must be detected and monitored by physicians. In clinical routing, manual 2D measurements are performed on 2D ultrasound (US) images to estimate the CVS volume but this practice is imprecise due to the unavailability of 3D information. A way to tackle this problem would be to develop automatic CVS segmentation algorithms for 3D US data. In this paper, we investigate the potential of 2D and 3D Convolutional Neural Networks (CNN) to solve this complex task and propose to use Compositional Pattern Producing Network (CPPN) to enable Fully Convolutional Networks (FCN) to learn CVS location. Our database was composed of 25 3D US volumes collected on 21 preterm nenonates at the age of 35.8±1.6 gestational weeks. We found that the CPPN enables to encode CVS location, which increases the accuracy of the CNNs when they have few layers. Accuracy of the 2D and 3D FCNs reached intraobserver variability (IOV) in the case of dilated ventricles with Dice of 0.893±0.008 and 0.886±0.004 respectively (IOV = 0.898±0.008) and with volume errors of 0.45±0.42 cm3 and 0.36±0.24 cm3 respectively (IOV = 0.41±0.05 cm3). 3D FCNs were more accurate than 2D FCNs in the case of normal ventricles with Dice of 0.797±0.041 against 0.776±0.038 (IOV = 0.816±0.009) and volume errors of 0.35±0.29 cm3 against 0.35±0.24 cm3 (IOV = 0.2±0.11 cm3). The best segmentation time of volumes of size 320×320×320 was obtained by a 2D FCN in 3.5±0.2 s.

Spinal contribution to neuromuscular recovery differs between elbow-flexor and knee-extensor muscles after a maximal sustained fatiguing task.

Data from studies of elbow-flexor (EF) or knee-extensor (KE) muscles suggest that a fatigue-related decrease in motoneuron excitability only occurs in EF. It is unknown how motoneuron excitability changes after sustained fatiguing maximal voluntary isometric contractions (MVICs) in EF and KE in the same participants. In two sessions, eight healthy men performed a 2-min MVIC of EF or KE to induce fatigue with brief MVICs before and six times after the 2-min MVIC. Electromyographic responses elicited by corticospinal tract stimulation at the transmastoid [cervicomedullary motor-evoked potential (CMEP)] or thoracic [thoracic motor-evoked potential (TMEP)] level were recorded from EF and KE, respectively. To account for muscle excitability, CMEPs and TMEPs were normalized to maximal M-wave (Mmax) elicited by peripheral nerve stimulation during each brief MVIC. Immediately after the 2-min MVIC, biceps brachii and brachioradialis CMEP/Mmax were 88% (SD 11%) (P = 0.026) and 87% (SD 12%) (P = 0.029) of pre-MVIC (PRE) values, respectively, and remained lower than PRE after 5 s of recovery [91% (SD 8%), P = 0.036 and 87% (SD 13%), P = 0.046, respectively]. No subsequent time points differed from PRE (all P ≥ 0.253). TMEP/Mmax for rectus femoris and vastus lateralis were not different from PRE at any time during the recovery period (all P > 0.050). A different recovery pattern in motoneuron excitability occurred in EF as it recovered by 60 s whereas KE motoneurons were unaffected by the fatiguing task. The present findings may contribute to better understand muscle-specific neurophysiological differences in spinal excitability.NEW & NOTEWORTHY By comparing the changes in motoneuron excitability in elbow-flexor and knee-extensor muscles after sustained fatiguing maximal voluntary contractions, this study shows that motoneuron recovery behavior depends on the muscle performing the exercise. A different recovery pattern in motoneuron excitability occurs in elbow flexors as it recovered by 60 s whereas knee extensors were unaffected by fatigue. This finding can help to increase understanding of the effect of a fatigue and subsequent recovery on neural processes.

Sustained Maximal Voluntary Contractions Elicit Different Neurophysiological Responses in Upper- and Lower-Limb Muscles in Men.

This study compared the effects of fatigue on corticospinal responsiveness in the upper- and lower-limb muscles of the same participants. Seven healthy males performed a 2-min maximal voluntary isometric contraction of the elbow flexors or knee extensors on four separate days. Electromyographic responses were elicited by nerve stimulation (maximal M-wave) in all sessions and by transcranial magnetic stimulation (motor-evoked potential; silent period) and spinal tract stimulation (cervicomedullary or thoracic motor-evoked potentials; silent period) in one session each per limb. During sustained maximal voluntary contractions, motor-evoked potential area normalised to M-waves increased from baseline in biceps brachii (155 ±â€¯55%) and rectus femoris (151 ±â€¯44%) (both p ≤ 0.045). At the end of maximal voluntary contractions, spinal tract motor-evoked potential area normalised to M-waves was smaller than baseline in biceps brachii (74 ±â€¯23%; p = 0.012) but not rectus femoris (108 ±â€¯40%; p = 0.999). The ratio of motor-evoked potential to spinal tract-evoked potential areas increased dramatically from 90 to 115 s in biceps brachii (p = 0.001) but not in rectus femoris (p = 0.999). Silent period durations increased similarly in both muscles (p ≤ 0.008) after transcranial and spinal stimulation. Sustained maximal contractions elicit different neurophysiological adjustments in upper- and lower-limb muscles. Specifically, motoneuronal excitability was reduced in biceps brachii, but not in rectus femoris, and this reduction required greater compensatory adjustments from the motor cortex. Therefore, changes in cortical and spinal excitability during sustained maximal exercise are likely specific to the muscle performing the task.

Sequential Emboli Detection From Ultrasound Outpatient Data.

This paper addresses the detection of emboli from signals acquired with a new miniaturized and portable transcranial Doppler ultrasound device. The use of this device enables outpatient monitoring but increases the number of artifacts. These artifacts usually come from the patient voice and motion and can be superimposed to emboli. For this reason and because of the scarcity of emboli compared to artifacts, reliably detect emboli is a challenging task. As an example, the 11809 s of signal used in this study contained 0.06 % of embolic events and 10.14 % of artifacts. Herein, we propose an automatic and sequential approach. The method is based on sequential determination of high intensity transient signals. We also define efficient features to describe emboli in the time frequency representation. On our database, the number of artifacts detected as emboli is divided by more than 10 compared to the other algorithms reported in the literature.

Do aerobic characteristics explain isometric exercise-induced neuromuscular fatigue and recovery in upper and lower limbs?

This study investigated the relationships between aerobic characteristics and (i) neuromuscular fatigue induced by 2-min sustained isometric maximal voluntary contractions (MVC) and (ii) subsequent recovery, in the upper and lower limbs. In a pseudo-randomized order, eleven healthy males completed four sessions on different days: maximal incremental cycling test (100 W + 40 W every 2 min); maximal arm-cranking test (50 W + 20 W every 2 min); and 2-min sustained isometric MVCs of the knee extensors (KE) and elbow flexors (EF). Neuromuscular assessment was performed with transcranial magnetic and peripheral nerve stimulation to evaluate central and peripheral neuromuscular factors of fatigue and the subsequent recovery. Peak oxygen uptake, gas exchange threshold and the corresponding power outputs were correlated with recovery of voluntary force after the 2-min KE MVC. Regression analysis showed that power output at the gas exchange threshold alone explained 72% of the variability in ∆recovery of KE voluntary force. No relationships with fatigue or recovery in EF were observed. These results suggest that participants with greater aerobic capacities experience the same amount of fatigue and faster recovery of voluntary force in KE but not EF. The potential reasons behind the relationship in KE but not EF are discussed.

Mechanisms of Fatigue and Recovery in Upper versus Lower Limbs in Men.

PURPOSE: To compare the mechanisms of fatigue and recovery between upper and lower limbs in the same subjects. METHODS: Twelve healthy young men performed a 2-min sustained maximal voluntary isometric contraction (MVC) of the knee extensors (KE) and on another day a 2-min MVC of the elbow flexors (EF). Neuromuscular function evaluations were performed with both transcranial magnetic and peripheral stimulations before (PRE), at the end of the 2-min MVC, and five more times within 8 min of recovery. RESULTS: Decreases in MVC and cortical voluntary activation were approximately 12% (P < 0.001) and approximately 25% greater (P = 0.04) in KE than EF at end of the 2-min MVC. Conversely, twitch response decreased approximately 29% more (P = 0.02) in EF than KE. Changes in motor-evoked potential with fatigue were not different between upper and lower limbs (P > 0.05), whereas the increase in silent period duration was approximately 30% greater in EF than KE (P < 0.05). CONCLUSIONS: Upper and lower limbs presented different magnitudes of total, central and peripheral fatigue. Total neuromuscular fatigue and central fatigue were greater in KE than EF. Conversely, peripheral fatigue and corticospinal inhibition were greater in EF than KE.

Photofocusing: Light and flow of phototactic microswimmer suspension.

We explore in this paper the phenomenon of photofocusing: a coupling between flow vorticity and biased swimming of microalgae toward a light source that produces a focusing of the microswimmer suspension. We combine experiments that investigate the stationary state of this phenomenon as well as the transition regime with analytical and numerical modeling. We show that the experimentally observed scalings on the width of the focalized region and the establishment length as a function of the flow velocity are well described by a simple theoretical model.

Iterative bandgap engineering at selected areas of quantum semiconductor wafers.

We report on the application of a laser rapid thermal annealing technique for iterative bandgap engineering at selected areas of quantum semiconductor wafers. The approach takes advantage of the quantum well intermixing (QWI) effect for achieving targeted values of the bandgap in a series of small annealing steps. Each QWI step is monitored by collecting a photoluminescence map and, consequently, choosing the annealing strategy of the next step. An array of eight sites, 280 mum in diameter, each emitting at 1480 nm, has been fabricated with a spectral accuracy of better than 2 nm in a standard InGaAs/InGaAsP QW heterostructure that originally emitted at 1550 nm.

Scanning near-field optical microscopy as a tool for the characterization of multimode interference devices.

We report the scanning near-field optical microscopy (SNOM) characterization of a 4 x 4 multimode interference (MMI) device working at a wavelength of 1.55 microm and designed for astronomical signal recombination. A comprehensive analysis of the mapped propagating field is presented. We compare SNOM measurements with beam-propagation-method simulations and thus are able to determine the MMI structure's refractive-index contrast and show that the measured value is higher than the expected value. Further investigation allows us to demonstrate that good care must be taken with the refractive-index profile used in simulation when one deals with low-index contrast structures. We show evidence that a step-index contrast is not suitable for adequate simulation of our structure and present a model that permits good agreement between measured and simulated propagating fields.