Research
My main research interests focus on combining computational and engineering tools with neuroscientific knowledge to develop neuro-technologies that can improve the quality of life of people with disability. I have participated in several projects that focused on creating neural interfaces and testing their feasibility with stroke and spinal cord injury patients.
Hybrid brain-machine interfaces
In the RamosLab, we worked on the development of hybrid brain-machine interfaces to improve stroke rehabilitation. The brain activity, measured with EEG, can be combined with residual muscle activity, measured with EMG, to exploit the latent capabilities of each patient.
We proposed a biologically-inspired hybrid brain-machine interface for upper-limb motor rehabilitation of stroke patients. Our approach detects the intention to move of patients with severe upper-limb paralysis by analyzing their cortical activity with EEG. Once the system detects that the patient is trying to move, the residual activity of the muscles is used to decode the exact movement that the patient is trying to perform, guiding the trajectory of the robotic exoskeleton. With this approach, the whole motor network is engaged, from the brain to the muscles. We hypothesize that the repetitive use of this system would facilitate motor recovery better than other approaches based on state-of-the-art neurotechnology.
Optimization of EEG-based detection of movement intentions in patients with motor deficits
Non-invasive brain-machine interfaces are a promising approach for future motor rehabilitation therapies. The intentions of movement of a paralyzed limb can be decoded with EEG and used to provide real-time proprioceptive feedback, which can facilitate motor recovery. One of the main challenges in this field is the accurate and reliable detection of the movement intentions. EEG measurements have several limitations, especially in terms of its low signal-to-noise ratio and easiness to get contaminated by artifacts. I have worked on several projects trying to exploit machine learning and signal processing strategies to optimize the information extracted from the EEG to detect the attempts of movement in patients with stroke and spinal cord injury.
Minimization of the influence of EEG artifacts
When patients try to move a paralyzed limb, they also generate compensatory movements with other parts of their body, contaminating the EEG activity. In this study, we quantified the impact of these EEG artifacts when stroke patients with complete hand paralysis tried to open and close their hand. Our results evidenced the relevance of artifact removal for the accurate estimation of movement information from EEG activity.
Exploiting information from previous recordings
One of the main limitations for the actual implantation of BMI therapies in clinical practice is the long set-up time. We proposed different recalibration strategies based on exploiting data recorded on previous days to maximize BMI decoding accuracy. We observed that combining data from past recordings with a short screening of less than 10 minutes is enough to achieve optimal performance.
Automatization of feature extraction
The brain of each person is unique and so is the activity that it generates during movement. To optimize the process of decoding movement intentions, we proposed an unsupervised strategy that extracts and selects the best features for each subject. It automatically captures the two EEG correlates of movement: the event-related desynchronization (ERD), and the movement-related cortical potentials (MRCP). This way, the BMI can be easily personalized for each patient. Our approach was evaluated to decode the initiation of seven different upper-limb movements and to detect movement attempts in spinal cord injury patients.
Brain-controlled functional-electrical stimulation for tetraplegic patients
In a joint cooperation between the University of Zaragoza (Spain), the Hospital Nacional de Parapléjicos (Toledo, Spain) and the CSIC (Madrid, Spain), we developed a brain-machine interface to control electrical stimulation of the paralyzed hand muscles in patients with tetraplegia.
Patients trained with the system for five consecutive days, controlling the muscles of one of their hands with their movement intentions decoded with EEG. After five days of training, the stimulated hand showed higher improvements in clinical scores, compared with the non-stimulated hand. Patients also reported high values of usability and satisfaction with the system.
EEG-control of an ambulatory exoskeleton for spinal cord injury gait rehabilitation
In this project, the University of Zaragoza (Spain), the Hospital Nacional de Parapléjicos (Toledo, Spain) and the Institute for Bioengineering of Catalonia (IBEC, Barcelona, Spain) joined forces to develop the first brain-controlled ambulatory exoskeleton (i.e., without any weight or balance support) for patients with paraplegia.
We proposed a protocol with special emphasis on safety, since patients with poor balance were required to stand and walk. The system was validated with four patients with spinal cord injury, who could successfully control the robotic exoskeleton with their brain activity. Patients reported low exertion and fatigue during the use of the system, and positive usability and satisfaction scores.
Characterization and restoration of motor cortical activation after spinal cord injury
The main consequence of a spinal cord injury (SCI) is the loss of motor and sensory function caudal to the level of injury. However, an SCI also results in a progressive brain reorganization as a consequence of the deafferentation. In order to develop neuro-technologies that work for these patients, it is important to characterize the cortical reorganization happening during the first months after the injury. Understanding this reorganization is the first step towards developing interventions that can revert the process, and restore the cortical function even years after SCI.
Evolution of EEG activity after SCI
In a longitudinal study, we characterized the progressive loss of event-related desynchronization (ERD) when patients with tetraplegia attempted to move their completely paralyzed hand, or when they imagined this movement. Eighteen patients were recruited shortly after their injury, and measured every two weeks during at least two months. We observed that the changes in motor cortical activation were correlated with the clinical progression of the patients.
Restoration of the motor cortical activation with EEG neurofeedback
Brain-machine interfaces for rehabilitation link the brain activation occurring during the attempt of movement with peripheral feedback. They generally assume that patients with SCI have intact brain activity. As observed in the previous study, the cortical activation measured in the EEG as the event-related desynchronization (ERD) decreases with time, and some chronic SCI patients do not show any measurable activity during the attempt of movement of their paralyzed limbs. In this study we proposed a novel intervention to restore motor cortical activation in chronic SCI patients. Based on previous evidence showing that higher tonic EEG alpha power is associated with higher ERD, we hypothesized that artificially increasing the alpha power over the motor cortex of these patients could enhance their ERD (i.e., motor cortical activation) during movement attempts. We used EEG neurofeedback (NF) to enhance the tonic EEG alpha power, providing real-time visual feedback of the alpha oscillations measured over the motor cortex. This approach was evaluated in a C4, ASIA A, SCI patient (9 months after the injury) who did not present ERD during the movement attempts of his paralyzed hands. The patient performed four NF sessions (in four consecutive days) and screenings of his EEG activity before and after each session. After the intervention, the patient presented a significant increase in the alpha power over the motor cortex, and a significant enhancement of the mu ERD in the contralateral motor cortex when he attempted to close the assessed right hand.