Atilla Kilicarslan

High accuracy decoding of user intentions using EEG to control a lower-body exoskeleton

Brain-Machine Interface (BMI) systems allow users to control external mechanical systems using their thoughts. Commonly used in literature are invasive techniques to acquire brain signals and decode users attempted motions to drive these systems (e.g. a robotic manipulator). In this work we use a lower-body exoskeleton and measure the users brain activity using non-invasive electroencephalography (EEG). The main focus of this study is to decode a paraplegic subjects motion intentions and provide him with the ability of walking with a lower-body exoskeleton accordingly. We present our novel method of decoding with high offline evaluation accuracies (around 98%), our closed loop implementation structure with considerably short on-site training time (around 38 sec), and preliminary results from the real-time closed loop implementation (NeuroRex) with a paraplegic test subject.

Sitting and standing intention can be decoded from scalp EEG recorded prior to movement execution

Low frequency signals recorded from non-invasive electroencephalography (EEG), in particular movement-related cortical potentials (MRPs), are associated with preparation and execution of movement and thus present a target for use in brain-machine interfaces. We investigated the ability to decode movement intent from delta-band (0.1–4 Hz) EEG recorded immediately before movement execution in healthy volunteers. We used data from epochs starting 1.5 s before movement onset to classify future movements into one of three classes: stand-up, sit-down, or quiet. We assessed classification accuracy in both externally triggered and self-paced paradigms. Movement onset was determined from electromyography (EMG) recordings synchronized with EEG signals. We employed an artifact subspace reconstruction (ASR) algorithm to eliminate high amplitude noise before building our time-embedded EEG features. We applied local Fishers discriminant analysis to reduce the dimensionality of our spatio-temporal features and subsequently used a Gaussian mixture model classifier for our three class problem. Our results demonstrate significantly better than chance classification accuracy (chance level = 33.3%) for the self-initiated (78.0 ± 2.6%) and triggered (74.7 ± 5.7%) paradigms. Surprisingly, we found no significant difference in classification accuracy between the self-paced and cued paradigms when using the full set of non-peripheral electrodes. However, accuracy was significantly increased for self-paced movements when only electrodes over the primary motor area were used. Overall, this study demonstrates that delta-band EEG recorded immediately before movement carries discriminative information regarding movement type. Our results suggest that EEG-based classifiers could improve lower-limb neuroprostheses and neurorehabilitation techniques by providing earlier detection of movement intent, which could be used in robot-assisted strategies for motor training and recovery of function.

A robust adaptive denoising framework for real-time artifact removal in scalp EEG measurements.

OBJECTIVE Non-invasive measurement of human neural activity based on the scalp electroencephalogram (EEG) allows for the development of biomedical devices that interface with the nervous system for scientific, diagnostic, therapeutic, or restorative purposes. However, EEG recordings are often considered as prone to physiological and non-physiological artifacts of different types and frequency characteristics. Among them, ocular artifacts and signal drifts represent major sources of EEG contamination, particularly in real-time closed-loop brain-machine interface (BMI) applications, which require effective handling of these artifacts across sessions and in natural settings. APPROACH We extend the usage of a robust adaptive noise cancelling (ANC) scheme ([Formula: see text] filtering) for removal of eye blinks, eye motions, amplitude drifts and recording biases simultaneously. We also characterize the volume conduction, by estimating the signal propagation levels across all EEG scalp recording areas due to ocular artifact generators. We find that the amplitude and spatial distribution of ocular artifacts vary greatly depending on the electrode location. Therefore, fixed filtering parameters for all recording areas would naturally hinder the true overall performance of an ANC scheme for artifact removal. We treat each electrode as a separate sub-system to be filtered, and without the loss of generality, they are assumed to be uncorrelated and uncoupled. MAIN RESULTS Our results show over 95-99.9% correlation between the raw and processed signals at non-ocular artifact regions, and depending on the contamination profile, 40-70% correlation when ocular artifacts are dominant. We also compare our results with the offline independent component analysis and artifact subspace reconstruction methods, and show that some local quantities are handled better by our sample-adaptive real-time framework. Decoding performance is also compared with multi-day experimental data from 2 subjects, totaling 19 sessions, with and without [Formula: see text] filtering of the raw data. SIGNIFICANCE The proposed method allows real-time adaptive artifact removal for EEG-based closed-loop BMI applications and mobile EEG studies in general, thereby increasing the range of tasks that can be studied in action and context while reducing the need for discarding data due to artifacts. Significant increase in decoding performances also justify the effectiveness of the method to be used in real-time closed-loop BMI applications.

Real-time strap pressure sensor system for powered exoskeletons.

Assistive and rehabilitative powered exoskeletons for spinal cord injury (SCI) and stroke subjects have recently reached the clinic. Proper tension and joint alignment are critical to ensuring safety. Challenges still exist in adjustment and fitting, with most current systems depending on personnel experience for appropriate individual fastening. Paraplegia and tetraplegia patients using these devices have impaired sensation and cannot signal if straps are uncomfortable or painful. Excessive pressure and blood-flow restriction can lead to skin ulcers, necrotic tissue and infections. Tension must be just enough to prevent slipping and maintain posture. Research in pressure dynamics is extensive for wheelchairs and mattresses, but little research has been done on exoskeleton straps. We present a system to monitor pressure exerted by physical human-machine interfaces and provide data about levels of skin/body pressure in fastening straps. The system consists of sensing arrays, signal processing hardware with wireless transmission, and an interactive GUI. For validation, a lower-body powered exoskeleton carrying the full weight of users was used. Experimental trials were conducted with one SCI and one able-bodied subject. The system can help prevent skin injuries related to excessive pressure in mobility-impaired patients using powered exoskeletons, supporting functionality, independence and better overall quality of life.

Simultaneous Scalp Electroencephalography (EEG), Electromyography (EMG), and Whole-body Segmental Inertial Recording for Multi-modal Neural Decoding

Recent studies support the involvement of supraspinal networks in control of bipedal human walking. Part of this evidence encompasses studies, including our previous work, demonstrating that gait kinematics and limb coordination during treadmill walking can be inferred from the scalp electroencephalogram (EEG) with reasonably high decoding accuracies. These results provide impetus for development of non-invasive brain-machine-interface (BMI) systems for use in restoration and/or augmentation of gait- a primary goal of rehabilitation research. To date, studies examining EEG decoding of activity during gait have been limited to treadmill walking in a controlled environment. However, to be practically viable a BMI system must be applicable for use in everyday locomotor tasks such as over ground walking and turning. Here, we present a novel protocol for non-invasive collection of brain activity (EEG), muscle activity (electromyography (EMG)), and whole-body kinematic data (head, torso, and limb trajectories) during both treadmill and over ground walking tasks. By collecting these data in the uncontrolled environment insight can be gained regarding the feasibility of decoding unconstrained gait and surface EMG from scalp EEG.

ANFIS based modeling and inverse control of a thin SMA wire

In this work, we propose an Adaptive Neuro Fuzzy Inference System (ANFIS) based hysteresis modeling and control strategy for a thin Shape Memory Alloy (SMA) wire. Controlling the SMA wire is a challenging problem because of its dynamic hysteretic behavior. By using a hybrid learning procedure ANFIS architectures are powerful tools for many applications, such as identifying nonlinear parameters in a controlled system, predicting chaotic time series and modeling nonlinear functions. We tested our ANFIS model by making it predict major and minor hysteresis loops in different driving frequencies and compared them with the experimental data. To compensate the hysteretic effect, we used an inverse ANFIS model and used it directly as a controller. After dramatically reducing the hysteretic effect, we implemented a PI control to fine tune the response.

Multiple kernel based region importance learning for neural classification of gait states from EEG signals

With the development of Brain Machine Interface (BMI) systems, people with motor disabilities are able to control external devices to help them restore movement abilities. Longitudinal validation of these systems is critical not only to assess long-term performance reliability but also to investigate adaptations in electrocortical patterns due to learning to use the BMI system. In this paper, we decode the patterns of users intended gait states (e.g., stop, walk, turn left, and turn right) from scalp electroencephalography (EEG) signals and simultaneously learn the relative importance of different brain areas by using the multiple kernel learning (MKL) algorithm. The region of importance (ROI) is identified during training the MKL for classification. The efficacy of the proposed method is validated by classifying different movement intentions from two subjects—an able-bodied and a spinal cord injury (SCI) subject. The preliminary results demonstrate that frontal and fronto-central regions are the most important regions for the tested subjects performing gait movements, which is consistent with the brain regions hypothesized to be involved in the control of lower-limb movements. However, we observed some regional changes comparing the able-bodied and the SCI subject. Moreover, in the longitudinal experiments, our findings exhibit the cortical plasticity triggered by the BMI use, as the classification accuracy and the weights for important regions—in sensor space—generally increased, as the user learned to control the exoskeleton for movement over multiple sessions.

Classification of stand-to-sit and sit-to-stand movement from low frequency EEG with locality preserving dimensionality reduction

Recent studies have demonstrated decoding of lower extremity limb kinematics from noninvasive electroencephalography (EEG), showing feasibility for development of an EEG-based brain-machine interface (BMI) to restore mobility following paralysis. Here, we present a new technique that preserves the statistical richness of EEG data to classify movement state from time-embedded low frequency EEG signals. We tested this new classifier, using cross-validation procedures, during sit-to-stand and stand-to-sit activity in 10 subjects and found decoding accuracy of greater than 95% on average. These results suggest that this classification technique could be used in a BMI system that, when combined with a robotic exoskeleton, can restore functional movement to individuals with paralysis.

Modeling and Hysteresis Compensation in a Thin SMA Wire Using ANFIS Methods

Having a responsive behavior to applied voltage/current, shape memory alloy (SMA) wires are widely used as miniature, low mass, compact actuators for many applications. Due to the non-linear hysteretic behavior, SMA wires require advanced control techniques for precision control when used as actuators. In this article, an adaptive neuro-fuzzy inference system (ANFIS) is developed to compensate for the hysteretic non-linearity in a thin SMA wire. An experimental setup is designed to test the SMA wire, which can achieve an operating bandwidth of 1 Hz without an external cooling system. By using the data obtained from the experimental system, an ANFIS model is developed to model the major and minor hysteresis loops (voltage vs. displacement characteristics) and the hysteresis non-linearity at various driving frequencies. Results of numerical simulations based on the ANFIS model closely match the experimental data. Having demonstrated the ability of the ANFIS to account for the hysteretic behavior, an inverse ANFIS controller is devised to compensate for the hysteretic behavior. Finally a PI control action is included in the feedback loop to fine-tune the system response in real time.

LPV Gain Scheduling Control of Hysteresis on an SMA Wire System

In this work, a Linear Parameter-Varying (LPV) control method is used to compensate the hysteretic behavior of a Shape Memory Alloy (SMA) wire. Controller is implemented on an experimental system which consists of a pre-tension spring and a mass actuated with a thin SMA wire. The hysteretic characteristic of the SMA wire is modeled using the Preisach model and the model is verified both for the major and minor hysteresis loops. The small signal linear gain of the Preisach model is used as a scheduling stiffness variable. The parameter-dependent controller is scheduled based on the real time measurement of the stiffness variable. An H∞ controller is also synthesized by representing the hysteresis as a parametric uncertainty and comparisons are made with LPV gain scheduling controllers using similar weights for both controllers. Experimental trajectory tracking results show that the LPV Gain Scheduling controller has a better response and the hysteresis uncertainty is compensated for the full range of stiffness variability.Copyright