Stroke Upper Limb Rehabilitation With Robot Operators: a Synergistic Approach for Human-Robot Interaction and Rehabilitative Trials.

The EU is under demographic aging with the median age of the population rising from 44.4 years in 2022 to an expected 48.8 years in 2070. This projection from the European Commission estimates that in 2070, there will be 1.7 persons over 65 years old for every person between the ages 20 to 64. This shift that is already taking place represents a major challenge to secure quality healthcare and long-term care to the population.

Italy's old-age dependency ratio is predicted to be among the highest ones across the EU. For this reason, a significant budget from Italy's National Recovery and Resilience Plan (PNRR) was attributed to improve hospitals and research in healthcare. Particularly, there is currently an emphasis in the research related to neurological-motor disorders caused by strokes, these being the third most common source of disability and death in Europe. Loss of autonomy due to brain injuries caused by a stroke can often remain permanent, requiring long-term support of patients.

The use of robots in rehabilitation is now a standard part of the treatment to restore motor functionality, complementing conventional physical therapy. Robots can provide benefits such as sensorimotor stimulation, assistive loads, and targeted, functional movements, all of which have the potential to improve treatment outcome and speed of recovery.

However, the mechanisms of motor recovery induced by robots for rehabilitation is still not clear and many studies are trying to define the optimal use of robots for therapy. Notably, it was found that an individualized selection of practice movements is beneficial to recovery. The role of assistance, resistance and error augmentation has been carefully studied to stimulate an improvement in movement coordination. Breaking down complex tasks into separate planar movements is also considered to enhance treatment.

These studies are able to orient clinicians in the choice of treatment, but there is still a lack of understanding of the physiological mechanisms underlying motor control and therefore of the effects of therapy on the neurological system. A deeper understanding of how muscular activity is controlled by the nervous system could help identify more clearly the mechanisms underlying a patient's disability and how to stimulate effective motor recovery. This would provide with a better assessment of a patient's condition and in turn help develop new and optimized treatments.

A long-standing theory of how the central nervous system (CNS) controls movements is that of muscle synergies. It is believed that the realization of any movement, involving the activation of thousands of neurons, stems from a hierarchical organization from the CNS. A relatively low number of neurological modules would control the simultaneous activation of many specific motor neurons to contract muscles synergistically.

Although there is no definitive proof of the existence of such modules in the spinal cord, many studies have identified the simultaneous co-activation of groups of muscles involved in specific motions. Correct reconstruction of muscle activation signals can be obtained from the low-dimensional information from synergies, revealing the possibility to encode complex movements into a combination of synergy modules. It is important to note that there are many different representations of such "modules", which can be time-invariant or variant as well as space-invariant or variant. This simple representation of muscle activation can be leveraged to analyze motor control, which would allow an assessment of motor functionality at a higher level. The study of muscle synergies in a clinical context could be a significant step forward to an individualized care of patients. It explores a different domain of movement compared to kinematic analysis, namely the modular structure of the neuromotor system, hence complementing the available set of information from a given patient. A notable difference between kinematics and synergistic analysis is the similarity across subjects: the information extracted from muscle synergies has strong similarity between identical subjects compared to kinematic data. This is an important factor to be able to assess a patient's relative condition.

In order to use muscle synergies as biomarkers in the clinic, the scientific literature is still missing a comprehensive analysis acknowledging the reliability of synergies to identify motor impairment, and to which extent it is able to do so. A few studies have started to answer this question by realizing experiments to analyze the stability of muscle synergies across patients, on the upper limb. It was found that synergies are a robust measure when considering variability between subjects for a given movement.

A recent study by Berger et al. has analyzed the effect of cerebellar ataxia on muscle synergies, concluding that the spatio-temporal organization of synergies is indeed affected by damage in the cerebellum. This study, along with others showcases a possible use of the synergistic model to assess a patient's level of damage and their recovery path. Following such results, studies have looked at the effect of rehabilitation and robotic assistance on muscle synergies of stroke-affected patients, either by extracting synergies during therapy or comparing pre- and post-therapy results.

There are still many open questions to solve before applying the method of muscle synergies to a clinical practice. The variables influencing the obtention of muscle synergies are still debated over in the literature: how synergies differ between patients and healthy controls, if there is a constant number of modules that should be used to represent the muscle signals, and the modalities to extract the synergies (number of EMG channels, muscles to consider, algorithms for dimensionality reduction and methods to process EMG data). These are important parameters to decide, which could be better resolved by assessing the reproducibility of muscle synergies of healthy controls between different sessions.

Another question relevant to clinical practice, is to understand how muscle synergies are affected by robotic devices typically used for therapy. Robot-assisted rehabilitation is characterized by two types: end-effector-type and exoskeleton-type rehabilitation robots. Both types showed to have promising results on rehabilitation, but few studies directly compared their effects on patients. A randomized trial analyzed the effects on motor recovery of end-effector robot and exoskeleton, finding greater improvements in the end-effector robotic rehabilitation. No study has yet identified synergy patterns from robotic use to characterize the human-robot interaction in post-stroke patients when using an end-effector robot or an exoskeleton robot. This comparison will be needed in the clinic to assess synergies during robotic rehabilitation as an assessment of patient's progress.

In clinical scenario, clinical scales and kinematic analysis are usually used for the evaluation of patients. However, these methods are limited, since they may suffer from the lack of objectivity and sensitivity. Synergies go beyond pure kinematics and allow to directly understand the neural mechanisms underlying the motor impairment. Synergy analysis has been exploited for understanding physiological mechanisms in healthy subjects, but only few studies analyzed muscle synergies on patients during rehabilitation. In order to translate this methodology to the clinical practice, there is the need of comprehensive and structured studies. However, only few pilot studies are available that analyzed synergies during rehabilitation. Thus, in this study we aimed at filling this gap by analyzing synergies during human-robot interaction when using an end-effector robot and an exoskeleton robot and by assessing the modifications of synergies after robotic rehabilitation in detail. This study is part of the PNRR30 project "Fit4MedRob-Fit for Medical Robotics" (Piano Nazionale Complementare (PNC)-PNC0000007), that aims at improving the current rehabilitation for patients with reduced or absent motor, by means of novel technologies and tools in all the phases of the rehabilitation process. The project includes novel algorithms and diagnostic tools that allow to improve the current rehabilitation protocols. Therefore, the present study will contribute to investigate the effects of robotic rehabilitation therapy on patients suffering from post-stroke motor impairment with the synergy methods. For this purpose, synergies will be analyzed with the SynMMF tool based on the novel mixed-matrix factorization (MMF) and on the SynergyAnalyzer Toolbox that extracts muscle synergies, kinematic-muscular synergies, functional synergies, and synergies with negative weights from EMG and multi-domain data. First, a single-session study will be conducted to evaluate the human-robot interaction, comparing the interaction of post-stroke patients with an upper limb exoskeleton and an upper limb end-effector robot. Then, a longitudinal study will be conducted in which patients are evaluated before and after rehabilitation by investigating the changes in synergies in a wide variety of movements.

This study may deepen the understanding of the mechanisms of motor control while interacting with robotic devices and help to understand better the neural mechanism underlying motor impairment and to improve the motor recovery. We develop in the next section a specific set of hypotheses and objectives that the study will answer.