Motor Imagery (MI)-based Brain–Computer Interfaces (BCIs) have been widely used as an alternative communication channel to patients with severe motor disabilities, achieving high classification accuracy through machine learning techniques. Recently, deep learning techniques have spotlighted the state-of-the-art of MI-based BCIs. These techniques still lack strategies to quantify predictive uncertainty and may produce overconfident predictions. In this work, methods to enhance the performance of existing MI-based BCIs are proposed in order to obtain a more reliable system for real application scenarios. First, the Monte Carlo dropout (MCD) method is proposed on MI deep neural models to improve classification and provide uncertainty estimation. This approach was implemented using Shallow Convolutional Neural Network (SCNN-MCD) and with an ensemble model (E-SCNN-MCD). As another contribution, to discriminate MI task predictions of high uncertainty, a threshold approach is introduced and tested for both SCNN-MCD and E-SCNN-MCD approaches. The BCI Competition IV Databases 2a and 2b were used to evaluate the proposed methods for both subject-specific and non-subject-specific strategies, obtaining encouraging results for MI recognition.

Many studies applying Brain-Computer Interfaces (BCIs) based on Motor Imagery (MI) tasks for rehabilitation have demonstrated the important role of detecting the Event-Related Desynchronization (ERD) to recognize the user's motor intention. Nowadays, the development of MI-based BCI approaches without or with very few calibration stages session-by-session for different days or weeks is still an open and emergent scope. In this work, a new scheme is proposed by applying Convolutional Neural Networks (CNN) for MI classification, using an end-to-end Shallow architecture that contains two convolutional layers for temporal and spatial feature extraction. We hypothesize that a BCI designed for capturing event-related desynchronization/synchronization (ERD/ERS) at the CNN input, with an adequate network design, may enhance the MI classification with fewer calibration stages. The proposed system using the same architecture was tested on three public datasets through multiple experiments, including both subject-specific and non-subject-specific training. Comparable and also superior results with respect to the state-of-the-art were obtained. On subjects whose EEG data were never used in the training process, our scheme also achieved promising results with respect to existing non-subject-specific BCIs, which shows greater progress in facilitating clinical applications.

The integration of lower limb exoskeletons with robotic walkers allows obtaining a system to improve mobility and security during gait rehabilitation. In this work, the evaluation of human motion intention (HMI) based on electroencephalogram (EEG) and surface electromyography (sEMG) signals are analyzed for a knee exoskeleton control as a preliminary study for gait neuro-rehabilitation with a hybrid robotic system. This system consists of the knee exoskeleton H2 and the UFES's Smart Walker, which are used to restore the neuromotor control function of subjects with neural injuries. An experimental protocol was developed to identify patterns to control the exoskeleton in accordance with the HMI-based on EEG/sEMG. The EEG and sEMG signals are recorded during the following activities: stand-up/sit-down and knee flexion/extension. HMI is analyzed through both event-related desynchronization/synchronization (ERD/ERS) and slow cortical potential, as well as the myoelectric pattern classification related to lower limb. The feature extraction from sEMG signals is based on vector combinations in time and frequency domain which are used for a pattern classification stage trough an artificial neural network with Levenberg Marquadt training algorithm and support vector machine. Preliminary results shown that a combination of EEG/sEMG signals can be used to define a control strategy for the robotic system

Backgrounds: Exoskeletons development arises with a leading role in neurorehabilitation technologies; however, very few prototypes for upper limbs have been tested, contrasted and duly certified in terms of their effectiveness in clinical environments in order to incorporate into the health system. The purpose of this pilot study was to determine if robotic therapy of Hemiplegic Shoulder Pain (HSP) could lead to functional improvement in terms of diminishing of pain, spasticity, subluxation, the increasing of tone and muscle strength, and the satisfaction degree. Methods: An experimental study was conducted in 16 patients with painful shoulder post- ischemic stroke in two experimental groups: conventional and robotic therapy. At different stages of its evolution, the robotic therapy effectiveness applied with anti-gravitational movements was evaluated. Clinical trial was developed at the Physical Medicine and Rehabilitation Department of the Surgical Clinical Hospital “Dr. Juan Bruno Zayas Alfonso” in Santiago de Cuba, from September 2016 - March 2018. Among other variables: the presence of humeral scapular subluxation (HSS), pain, spasticity, mobility, tone and muscle strength, and the satisfaction degree were recorded. Results with 95% reliability were compared between admission and third months of treatment. The Mann-Whitney U-Test, Chi-Square and Fisher’s Exact Tests were used as comparison criteria. Results: Robotic therapy positively influenced in the decrease and annulment of pain and the spasticity degree, reaching a range increase of joint movement and the improvement of muscle tone.

The development of Brain–Computer Interfaces based on Motor Imagery (MI) tasks is a relevant research topic worldwide. The design of accurate and reliable BCI systems remains a challenge, mainly in terms of increasing performance and usability. Classifiers based on Bayesian Neural Networks are proposed in this work by using the variational inference, aiming to analyze the uncertainty during the MI prediction. An adaptive threshold scheme is proposed here for MI classification with a reject option, and its performance on both datasets 2a and 2b from BCI Competition IV is compared with other approaches based on thresholds. The results using subject-specific and non-subject-specific training strategies are encouraging. From the uncertainty analysis, considerations for reducing computational cost are proposed for future work.