Titolo della tesi: Re-establishing Cortico-Muscular Communication to enhance recovery: development of an hybrid Brain-Computer Interface for post-stroke motor rehabilitation
Stroke is the leading cause of adult long-term disability in Western countries and the second leading cause of death worldwide. The most common and widely
recognized impairment caused by stroke is motor impairment contralateral to the affected brain hemisphere (hemiparesis). Notably, the main predictor of
an individual resuming a normal professional and personal life is upper limb extremity function. Indeed, improving upper limb functioning is the primary therapeutic goal in stroke rehabilitation to maximize patients’ functional recovery and reduce long-term disability.
Various innovative neurorehabilitation strategies are emerging in order to enhance beneficial plasticity and improve motor recovery after stroke. Among
them, Brain-Computer Interfaces (BCIs) have proven their efficacy to enhance upper limb motor recovery exploiting brain signals to control visual or proprioceptive feedbacks/effectors. As for the feedback, Functional Electrical Stimulation (FES) has been employed in stroke rehabilitation for its capability
to assist movement and has been shown to induce changes in the brain, bearing witness of brain plasticity modulation.
BCIs for post-stroke motor rehabilitation rely on the principle that reinforcement of close-to-normal motor related brain activity (most commonly
derived from electroencephalogram - EEG), results in an improvement of motor function. However, along the process of motor recovery after stroke, several
abnormalities in upper limb function have been described such as muscle weakness and spasticity, abnormal muscle co-activation, increased activity of the
antagonist muscles. Electromyography (EMG) can be used to monitor the residual or recovered muscular activity along the rehabilitation processes and EMG-related features can be exploited to avoid the reinforcement of such maladaptive changes. Hybrid BCIs (h-BCIs) include peripheral signals
such as EMG, in addition to brain signals, as control feature and they have mostly been developed to improve the classification performance of the
system as in assistive BCIs. Such devices usually combine the EEG and EMG feature in the classification stage, meaning that each feature (brain
and muscular) is calculated separately and combined sequentially or simultaneously using a balanced weight or Bayesian fusion approach to better control
the assistive device. Nevertheless, there is no consensus on which movement-related features should be encouraged (or discouraged) within a BCI
training to pursue physiological muscular activation patterns. Ideally, h-BCI systems specifically developed for hand motor rehabilitation should allow to
train both brain and peripheral activity in a top-down framework in which volition, that is brain control over muscular activation, is reinforced together
with correct muscular activation patterns.
Thus, here Cortico-Muscular coupling (CMC) is proposed as feature of a novel h-BCI system aimed at re-establishing the brain-muscles communication
after stroke. CMC gives information on how much cortical surface motor potentials are phase-locked to muscular firing during voluntary movement. It can
be considered a simple form of hybrid functional connectivity measuring the spectral coherence between EEG and EMG [29]. It has been proposed as a
potential biomarker for post-stroke motor deficits [30], indeed its amplitude has been proven to be reduced post-stroke and its increase has been correlated
with functional recovery. Recently, h-BCIs based on CMC have been studied for post-stroke motor rehabilitation. However, so far CMC studies have considered mainly single EEG-EMG channels combinations, disregarding the comprehensive functional connectivity pattern involving several
brain regions and muscles.
During the three years of my PhD, I implemented a non-invasive BCI-controlled FES device (RECOM) for upper limb rehabilitation after stroke
based on online detection of cortico-muscular activation. The control feature was derived from a combined EEG and EMG connectivity pattern estimated
during upper limb movement attempts. In particular, the first year was dedicated to the study of the state of the art and the development of a methodology
for the effective extraction of CMC patterns able to characterize physiological movements and to be used for movement classification. Moreover, to analyse
the functional connectivity between cortex and muscles after stroke, an ad-hoc protocol was developed for multimodal acquisition of stroke patients and, during the second year the data collected were used to characterize brain-muscles patterns during the movement of the impaired hand. Finally, the CMC computation was translated in real-time and the third year was dedicated to the design and the feasibility testing of a reliable and easy-to-use rehabilitative
h-BCI system based on CMC features. Moreover, a study on the strategy of the feedback delivery (i.e. FES) was performed with the ultimate aim of tailoring
the stimulation to patients’ impairment.