TYPE Methods

PUBLISHED 28 August 2025

DOI 10.3389/fbioe.2025.1655416

User-centered development of a
OPEN ACCESS personalized adaptive mirror
EDITED BY
Yu Cao,
University of Leeds, United Kingdom
therapy for upper-limb
REVIEWED BY
Xin Wang,
post-stroke rehabilitation using
The University of Leeds, United Kingdom
Jun Huo,
Huazhong University of Science and
virtual reality and myoelectric
Technology, China

*CORRESPONDENCE
control
Paolo De Pasquale,
paolo.depasquale@irccsme.it
Andrea d’Avella,
Paolo De Pasquale 1†*, Daniela De Bartolo 2†, Marta Russo 3,4,
a.davella@hsantalucia.it Denise J. Berger 3,5, Antonella Maselli 3,6, Daniele Borzelli 3,6,
†
These authors have contributed equally to this Emma Colamarino 7,8, Donatella Mattia 8, Christian Nissler 9,
work to the work
Markus Nowak 9, Elena Falomo 10, Javier Soto Morras 10,
RECEIVED 27 June 2025
ACCEPTED 31 July 2025 Moco Raffael Schiller 10, Claudio Castellini 9,11,
PUBLISHED 28 August 2025
Giovanni Morone 2,12 and Andrea d’Avella 3,13*
CITATION
1
De Pasquale P, De Bartolo D, Russo M, IRCCS Centro Neurolesi Bonino Pulejo, Messina, Italy, 2Clinical Laboratory of Experimental
Berger DJ, Maselli A, Borzelli D, Colamarino E, Neurorehabilitation, IRCCS Fondazione Santa Lucia, Rome, Italy, 3Laboratory of Neuromotor Physiology,
Mattia D, Nissler C, Nowak M, Falomo E, IRCCS Fondazione Santa Lucia, Rome, Italy, 4Institute of Cognitive Sciences and Technologies, Consiglio
Soto Morras J, Schiller MR, Castellini C, Nazionale delle Ricerche (CNR), Rome, Italy, 5Department of Systems Medicine and Center of Space
Morone G and d’Avella A (2025) User-centered Biomedicine, University of Rome Tor Vergata, Rome, Italy, 6Department of Biomedical Sciences,
development of a personalized adaptive mirror Dentistry and Morpho-Functional Imaging, University of Messina, Messina, Italy, 7Department of
therapy for upper-limb post-stroke Computer, Control and Management Engineering, Sapienza University of Rome, Rome, Italy,
8
rehabilitation using virtual reality and Neuroelectrical Imaging and Brain Computer Interface Lab, IRCCS Fondazione Santa Lucia, Rome, Italy,
9
myoelectric control. Institute of Robotics and Mechatronics, German Aerospace Center (DLR), Weßling, Germany, 10NEEEU
Front. Bioeng. Biotechnol. 13:1655416. Spaces GmbH, Berlin, Germany, 11Assistive Intelligent Robotics Lab, Friedrich-Alexander-Universität
doi: 10.3389/fbioe.2025.1655416 Erlangen-Nürnberg (FAU), Erlangen, Germany, 12Department of Life, Health and Environmental Sciences,
University of L’Aquila, L’Aquila, Italy, 13Department of Biology, University of Rome Tor Vergata, Rome, Italy
COPYRIGHT
© 2025 De Pasquale, De Bartolo, Russo, Berger,
Maselli, Borzelli, Colamarino, Mattia, Nissler,
Nowak, Falomo, Soto Morras, Schiller, Castellini,
Morone and d’Avella. This is an open-access
Introduction: Cerebral stroke often results in significant motor deficits, including
article distributed under the terms of the
Creative Commons Attribution License (CC BY). contralateral hemiparesis of the upper limb. Rehabilitation protocols with high-
The use, distribution or reproduction in other intensity and task-specific exercises can improve these deficits. Recent
forums is permitted, provided the original
technological advancements in virtual reality (VR), myoelectric control, and
author(s) and the copyright owner(s) are
credited and that the original publication in this exergames may be exploited to enhance rehabilitation effectiveness. However,
journal is cited, in accordance with accepted novel rehabilitation approaches combining these novel methodologies have
academic practice. No use, distribution or
rarely been developed with the active involvement of both therapists and patients.
reproduction is permitted which does not
comply with these terms.
Methods: An interdisciplinary team developed a novel system, Validation of the
Virtual Therapy Arm (VVITA), for post-stroke upper-limb rehabilitation combining
VR, myoelectric control, and exergames using a user-centered design (UCD)
approach. The VVITA hardware includes a head-mounted VR display, motion
tracking devices integrated in the VR system, and wireless armbands to record
electromyographic (EMG) signals, providing an interactive virtual environment for
immersive rehabilitation exercises implementing a virtual mirror therapy.
Assistance and task difficulty are adjusted dynamically based on patient
performance, promoting active participation and motor learning.

Frontiers in Bioengineering and Biotechnology 01 frontiersin.org

De Pasquale et al. 10.3389/fbioe.2025.1655416

Results: The development process involved iterative phases, involving focus groups
with stroke patients, therapists, and researchers. A pilot study with four stroke
survivors assessed the system’s feasibility, demonstrating its potential for
personalized and adaptive rehabilitation.

Conclusion: The VVITA system enhances mirror therapy by integrating VR and

myoelectric control, providing a tailored approach to upper-limb post-stroke
rehabilitation. The UCD approach ensured the system met patient and therapist
needs, showing promise for improving motor recovery and rehabilitation
outcomes.

KEYWORDS

stroke rehabilitation, virtual reality, myoelectric control, user-centered design,

mirror therapy

1 Introduction monitoring of the motor performance of the rehabilitation

exercise (De Pasquale et al., 2024). Affordable prices and ease of
Cerebral stroke is one of the main causes of disability in adults use make these systems suitable for home rehabilitation after
and most frequently involves hemiparesis of the contralateral body. hospitalization using a telemedicine approach (Piron et al., 2009).
Hemiparesis can cause muscular stiffness and other impairments to Myoelectric control is a novel and promising approach for
the fine and global motor coordination of the upper limbs, including rehabilitation. Myoelectric interfaces have been mostly used for
restricted range of motion of the arm’s joints which hinders reaching the control of actuators such as exoskeletons or prostheses
and grasping movements, essential to perform daily living activities (DiCicco et al., 2004; Liarokapis et al., 2013; Nissler et al., 2019).
(Hatem et al., 2016). Thanks to myoelectric interfaces, which decode the patient’s
These motor deficits occur for 80% of patients in the acute phase intention through the residual myoelectric activity in the paretic
and for 40% of patients in the chronic phase and can be improved limb, patients may generate voluntary movements through their
through rehabilitation protocols with high-intensity and/or task- spared cortico-spinal pathway and receive feedback (e.g., a realistic
specific exercises (Cramer et al., 1997; Kwakkel et al., 2004; Veerbeek visual feedback from an embodied limb in VR) thus establishing a
et al., 2014). Technological advances in recent years have provided closed-loop system that promotes re-learning and encourages active
new methodologies to support and foster the rehabilitation process participation, increasing motor coordination, muscle strength and
by increasing its repeatability and intensity (Foley et al., 2012; reducing spasticity (Song et al., 2013; Sarasola-Sanz et al., 2018). The
Cikajlo et al., 2020; Ceradini et al., 2024; De Luca et al., 2024), use of myoelectric interfaces for rehabilitation also aims at
including but not limited to virtual reality (VR), myoelectric control, promoting neuroplasticity to reshape neuromuscular activity and
and exergames. to enhance motor learning, and the restoration of motor function.
Rehabilitation systems integrated into VR systems can provide For instance, stroke patients can learn to control, with the more
training scenarios that are more complex and engaging than those affected upper limb, a multi-degree-of-freedom exoskeleton using a
employed in conventional therapy (CT) (Maggio et al., 2023). In decoder trained with EMG from the healthy limb (Sarasola-Sanz
addition, activities of daily living (ADL) can be simulated in an et al., 2022). Moreover, myoelectric control training can reduce
ecological and controlled manner thanks to the immersivity of head- abnormal co-activation (i.e., undesired coupling) by training only
mounted displays, large projection screens, and VR caves. Thus, VR the desired muscles while leaving other muscles unaffected (Seo
scenarios effectively increase patient’s active participation and et al., 2022).
motivation leading to a better adherence to the rehabilitation Virtual Reality and active video games that combine physical
protocol (Domínguez-Téllez et al., 2020). Immersive VR systems activity with interactive serious games, or exergames, have been used
currently available in consumer electronics and gaming markets can for rehabilitation (Maggio et al., 2020; Morone et al., 2024). In
also record movement kinematics and allow quantitative addition, a serious game was recently approved by the FDA (Food
and Drug Administration) as the first digital drug in children
affected by ADHD (Attention Disorder Hyperactivity Disorder)
(Laver et al., 2017; Commissioner, 2020; Mubin et al., 2022).
Abbreviations: VR, virtual reality; VITA, Virtual Therapy Arm; VVITA, Validation Exergames make rehabilitation exercises more enjoyable and
of the Virtual Therapy Arm; UCD, user-centered design; EMG,
electromyographic; CT, conventional therapy; ADL, activities of daily living; engaging. However, exergames commercially released often are
MT, mirror therapy; GUI, graphical user interface; VMAL, virtual more affected not tailored to the specific needs and requirements of patients
limb; RMAL, real more affected limb; RLAL, real less affected limb; USEQ, User and cannot easily incorporate feedback and suggestions from
Satisfaction Evaluation Questionnaire; VAS, Visual Analogical Scale; PPRS,
Pittsburgh Participation to rehabilitation Scale; NASA-TLX, NASA Task Load therapists. In contrast, recent advancements in assessment
Index; NASA-TLX, National Aeronautics and Space Administration Task Load methodologies can be exploited to devise personalized
Index; FMA, Fugl-Meyer Assessment; T0, beginning of the rehabilitation rehabilitation approaches, aimed at restoring specific components
protocol (baseline assessment); T1, end of the rehabilitation protocol
(post-treatment assessment); LMM, linear mixed model; α, agency of the motor deficits. For instance, such approaches may target the
parameter; β, capability parameter. motor functionality of the proximal upper limb (shoulder and elbow

Frontiers in Bioengineering and Biotechnology 02 frontiersin.org

De Pasquale et al. 10.3389/fbioe.2025.1655416

joints), the distal extremities (wrist and fingers), both and visual input (Giroux et al., 2018). Recent studies show that,
simultaneously, or they may focus on restoring balance or combining VR with MT leads to better rehabilitation results (Perez-
reducing treatment time (Henrique, Colussi, and De Marchi, Marcos et al., 2018), especially when using a contralateral lesion
2019). Customized approaches also allow to take into account action observation network (Saleh et al., 2017). Although it has been
patients’ needs and to modify the level of difficulty of the already proven that VR technology positively affects motor
exercise considering the patients’ functional assessment, ability functionality delivering highly immersive environments for motor
and capability, and the achieved level of motor recovery, which learning, the use of VR with MT protocols have so far shown limited
are not usually quantified in commercial game-based applications evidence of effectiveness due to the small number of recruited
(Tsekleves et al., 2016; Han et al., 2017; Triandafilou et al., 2018; patients, an inadequate research design and/or low-intensity
Zirbel et al., 2018). Customization also allows to directly involve training (Hsu et al., 2022).
therapists in the development of rehabilitation systems regarding the Therefore, we developed a novel system to overcome the
definition of system requirements including difficulty to use, time to limitations of a conventional MT by exploiting the potentiality of
set the system and the required knowledge (Rand et al., 2018). VR to enhance the effectiveness of rehabilitation. Specifically, the
Indeed, while VR, myoelectric control, and exergames represent system allows patients suffering from stroke upper-limb hemiparesis
promising methodologies that can be exploited to introduce novel to control the related virtual paretic arm thanks to an innovative
rehabilitation therapies, the active involvement of both therapists and control algorithm based on input from both limbs whose relative
patients during the whole development process is critical to improve contribution is modulated adaptively by monitoring the progress of
their usability and effectiveness. In fact, a user-centered design (UCD) patients. Here, we present the UCD of the system, from the initial
approach is increasingly used to develop, improve, and evaluate new concept that emerged within our interdisciplinary research team,
rehabilitation systems or procedures. This approach may also facilitate through a series of development and validation steps, up to the
system usage and integration in clinical and domestic settings thus definition of a specific system configuration and training protocols
overcoming the limitations of current approaches (Ríos-Hernández to be evaluated systematically in future studies.
et al., 2021; Semprini et al., 2022). Using the UCD approach, we developed the Validation of the
The goal of the present work is to present the development of a Virtual Therapy Arm (VVITA) system for upper-limb post-stroke
system combining VR and myoelectric control to implement a new rehabilitation starting from the existing Virtual Therapy Arm
mirror therapy (MT) approach for post-stroke upper-limb (VITA) system, designed for treating phantom-limb pain in
neuromotor rehabilitation. MT is a rehabilitation technique that people with limb-loss and performing prosthetic training (Nissler
has significant potential to be enhanced by exploiting new et al., 2019). The development process included focus group sessions
methodologies such as VR, myoelectric control, and personalized with clinical and motor control experts to ensure the translation of
exergames through a UCD development process. During engineering outcomes into clinical practice. Hence, the definition of
conventional MT patients watch their unaffected limb reflection in the new control modality of the system to be used with stroke
a mirror placed on a sagittal plane between patients’ limbs. The patients through intermediate assessments and, where needed, the
reflection of the unaffected limb in the mirror creates the illusion that refinement of the design. In summary, although previous studies
the affected limb moves effectively and painlessly and provides have explored VR-based mirror therapy and myoelectric interfaces
encouragement. MT was first developed to alleviate phantom limb separately or in combination, our system presents distinctive
discomfort in persons with amputation (Ramachandran and Rogers- innovations. First, we propose an adaptive bimanual control
Ramachandran, 1996), and then applied to stroke patients with algorithm that dynamically integrates EMG signals from both
weaker limbs to improve muscle control (Altschuler et al., 1999). upper limbs, modulating their relative contribution according to
According to a systematic review, MT is recommended as a valuable individual patient progress. This allows a personalized progression
approach to be integrated in the rehabilitation intervention of stroke of training intensity and supports active engagement of the paretic
patients (Hatem et al., 2016). However, conventional MT has a few limb beyond traditional unimanual or fixed-threshold approaches.
limitations that restrict its use in clinical settings including being Furthermore, the system was developed following a rigorous user-
monotonous, only providing a low-dose therapy, and requiring centered design process, involving clinicians and therapists
specialized equipment and a professional on-site (Horne et al., throughout all stages of development to ensure clinical relevance,
2015). Moreover, movements are limited by the physical usability, and effective integration into rehabilitation practice. These
dimension of the “mirror box”. By allowing for more clinically features represent a novel contribution to the field by addressing key
viable use of MT approach, VR-based therapy may overcome these limitations of conventional mirror therapy and previous VR-EMG
restrictions and motivate patients to perform rehabilitation protocol systems. Finally, we present the results obtained from a pilot study
including meaningful ADL tasks. VR is indeed seen as a potential involving four stroke survivors, providing an initial assessment of
method for delivering larger therapeutic dosages and enhancing post- the feasibility of our novel personalized adaptive mirror therapy
stroke arm/hand rehabilitation (Chen et al., 2015). approach for upper-limb post-stroke rehabilitation based on virtual
Arm-hand movement training in VR is a successful technique reality and myoelectric control.
for increasing the functional motor recovery of stroke patients,
thanks to VR granting the possibility to enhance multisensory
feedback in a highly controllable and versatile fashion (Laver 2 Methods
et al., 2017; Massetti et al., 2018). The mirror visual illusion that
appears in VR systems facilitates the multisensory integration by The VVITA system has been developed as a stroke rehabilitation
promoting the interaction between bilateral proprioceptive signals application of the VR platform originally developed for the

Frontiers in Bioengineering and Biotechnology 03 frontiersin.org

De Pasquale et al. 10.3389/fbioe.2025.1655416

treatment of phantom limb pain in upper-limb amputees within the 2.2 UCD approach
VITA project led by a research group at the Institute of Robotics and
Mechatronics of the German Aerospace Center (DLR) in Munich, A UCD approach was implemented in the VVITA project to
Germany. The VITA system originally developed for amputees has develop a novel application of the VITA system optimized for stroke
been modified to be used with post-stroke patients to rehabilitate the rehabilitation, involving a multidisciplinary team of biomedical
neuromotor functionality of the more affected upper limb. This engineers, motor control neuroscientists, neurologists, therapists,
development involved an interdisciplinary and international team and patients. A focus group, including seven stroke patients, one
composed of several research groups. The DLR group modified the physiotherapist, one psychologist, one specialist in physical
system according to suggestions provided by two groups at medicine and rehabilitation, one neurologist, five motor control
Fondazione Santa Lucia (FSL) in Rome, Italy, including motor scientists, and five engineers supported the protocol development.
control and rehabilitation experts. Following an iterative UCD This team brought expertise in neurological disorders, VR-based
approach, the team specified the requirements for new training therapies, and neuromotor rehabilitation to ensure the system met
protocols for upper-limb post-stroke neurorehabilitation, both clinical and patient needs. The technical development also
implemented those protocols in the new software named VVITA, involved NEEEU (NEEEU Spaces GmbH, Berlin, Germany), a
developed methods for their assessment, and performed several company focused on applying Human Centered Design processes
evaluations of the new system by testing it in a pilot study with to new technologies, which collaborated as subcontractor on the
stroke patients. design, qualitative research and development of the patient and
therapist journeys, the virtual environment and the digital therapy
platform for therapists. For these developments, we specifically used
2.1 The VITA system a service design–centered approach, leveraging tools such as journey
maps and service blueprints to map the user’s experience over time.
The VITA system (Nissler et al., 2019) is a low-cost VR By combining these methodologies, we ensured that the evolving
solution developed to treat phantom-limb pain of people with rehabilitation service was continuously refined in collaboration with
limb-loss or to perform prosthetic training. It uses an HTC Vive end users.
Pro VR platform (HTC Europe Co. Ltd., Slough, Berkshire, The development activities followed three iterative cycles, each
United Kingdom), including a head-mounted display and two composed of distinct phases (P). The first cycle included four phases:
trackers, and two Myo armbands (Thalmic Labs, Ontario, Canada) (0) evaluation of the system’s context of use, included only in first
with EMG sensors. One tracker and one armband are placed on the cycle, (1) specification of the rehabilitation protocol based on
each amputee’s arm. One of the Vive Trackers is placed on experts’ requirements, (2) software development to implement the
the unimpaired hand dorsum and provides the position and desired protocol, and (3) system evaluation with healthy participants
orientation of the hand. The second tracker is placed on the and stroke survivors. The initial phase (P0), focused on assessing the
remaining portion of the amputated limb to provide its position system’s intended context of use, defining scientific questions, and
and orientation. The Myo armband consists of eight bipolar identifying desired outcomes. Inputs from this phase guided the
surface EMG sensors measuring the activity of several forearm specification of system requirements and training protocols in the
muscles. The sensors are mechanically connected and arranged in subsequent phase (P1). The first version of the software was
an armband placed on the amputee’s forearms to record most developed in the third phase (P2) and it was critically tested and
of the muscle controlling opening and closing of the fingers. A evaluated during the fourth phase (P3). In the two additional cycles,
representation of the hands is provided in the virtual environment the first, second and third phases were repeated to refine the
as visual feedback. Virtual hand movements are directly predicted protocol, address updated system requirements, incorporate
from the kinematic recorded with the trackers while virtual feedback from therapists and patients, and resolve issues
finger movements are predicted from the forearm muscle identified during development (Figure 1).
signals using a model trained through a machine learning The first cycle (P0, P1, P2, P3), lasted 30 months and consisted of
procedure. The machine learning method, an iterative variant of 40 interactions, the second cycle (P4, P5, P6) lasted 2 months and
Random Fourier Features Ridge Regression (iRR-RFF) (Patel et al., consisted of 13 interactions, while the last cycle (P7, P8, P9) lasted
2017), trains a non-linear decoder mapping features from the EMG 4 months and consisted of 12 interactions (See Table 1).
signals onto finger gestures and used to control the feedback The final software release addressed all the issues that had
given by the virtual hands. The acquired signals (kinematic and emerged during the development, and it was tested in a pilot
EMG) are wirelessly transmitted via Bluetooth technology. Data study (P9) involving four stroke patients (two chronic and two
(kinematic and EMG signals) acquisition, model training, and sub-acute). This study evaluated the system’s feasibility,
prediction processes are performed by a laptop (Alienware m15, functionality, and usability through kinematic analysis and user
Dell) with a dedicated GPU (GeForce RTX 2060, Nvidia). The feedback, demonstrating the system’s potential for enhancing post-
virtual scene reproduces a house surrounded by nature (trees, lake, stroke rehabilitation.
mountains, etc.), in which various activities both inside the house
(cooking, playing the drums, stoking up a fireplace, etc.) and
external (picking fruit, etc.) can be carried out for rehabilitation 2.3 Incremental development and testing
purposes. Participants can navigate within the virtual environment
by moving to the desired area according to the exercise they choose The VVITA system was developed following a structured UCD
to perform. approach. Each phase was designed to iteratively refine the system to

Frontiers in Bioengineering and Biotechnology 04 frontiersin.org

De Pasquale et al. 10.3389/fbioe.2025.1655416

FIGURE 1
Schematic of the UCD process. The figure illustrates the iterative approach of the UCD methodology, detailing the various phases (P) that
transformed the initial VITA into the final VVITA system. The phases specify requirements, produce design solutions and evaluate designs were repeated
three times to refine the system and address issues identified during testing.

TABLE 1 Summary of UCD phases and cycles. The phases included in each cycle of the UCD approach: the type of phase, number of iterations conducted
within each phase, and a brief description of their objectives and outcomes.

Cycle Stage Type Number of interactions Description

1 P0 Meetings 7 Specify Context of Use

P1 Meetings 23 Specify Requirements

P2 Meetings 4 Produce design solution

P3 Testing healthy subjects 6 Evaluate designs

2 P4 Meetings 7 Specify Requirements

P5 Meetings 3 Produce design solution

P6 Pre-pilot stroke patients 3 Evaluate designs

3 P7 Meetings 6 Specify Requirements

P8 Meetings 2 Produce design solution

P9 Pilot stroke patients 4 Evaluate designs

meet the specific needs of post-stroke upper-limb rehabilitation. In botulinum toxin were excluded if such treatments occurred 2 weeks
this section we describe the main activities performed in each phase before or during rehabilitation. A comprehensive review of the
of development. The outcome of these activities, including the novel literature (Eng et al., 2007; Burke et al., 2009; Cheung et al.,
VVITA system and the results of the pilot study, are presented in 2009; Dohle et al., 2009; Bohil et al., 2011; Lee et al., 2011;
the Results. Thieme et al., 2012; Zhang and Zhou, 2012; Li et al., 2014;
Pollock et al., 2014; Ballester et al., 2015; Choi et al., 2016;
2.3.1 P0 – Specification of the context of use of the Hatem et al., 2016; Hoermann et al., 2017; Kim, 2017; Laver
VVITA system et al., 2017; Sadarangani et al., 2017; Darbois et al., 2018; Morone
The first phase defined the context of use, targeting subacute and et al., 2019) identified limitations in existing systems, such as
chronic stroke patients with diverse motor impairments. Inclusion inadequate adaptability and underutilization of immersive VR
criteria addressed motor deficits, age, and cognitive ability to ensure and myoelectric control. These findings guided the conceptual
compatibility with the system. Patients undergoing treatments like design of the VVITA system, which integrated VR mirror

Frontiers in Bioengineering and Biotechnology 05 frontiersin.org

De Pasquale et al. 10.3389/fbioe.2025.1655416

therapy, myoelectric control, and dynamic task adjustments to evaluation aimed to highlight limitations and challenges in
enhance patients’ engagement. Focus group discussions refined calibration and target placement procedures, as well as to identify
these concepts, emphasizing patient motivation, adaptive potential compensatory motor strategies exhibited by patients
protocols, and therapist control. These inputs formed the during tasks. Furthermore, the test supported therapists to
foundation for subsequent phases. identify the range of task parameters and type of feedback
mechanisms compatible with the intended applications.
2.3.2 P1 – Definition of the system requirements
and specifications 2.3.5 P4 – Refinement of the system requirements
Building on the outcomes of Phase P0, detailed requirements and specifications
were established for the hardware, software, and training protocols. Based on the findings from Phase P3, several refinements were
The hardware design included motion sensors for hands’ tracking implemented to optimize the system for usability and adaptability
and EMG sensor bracelets for myoelectric control. The software across diverse rehabilitation needs. Calibration procedures were
specifications encompassed immersive VR environments, enhanced to improve alignment accuracy, while target placement
calibration procedures, and assistive control algorithms for algorithms were revised to accommodate varying ranges of motion.
proximal (shoulder, elbow) and distal (hand) joints. Training Task parameters were adjusted to balance challenge and fatigue. For
protocols were designed to include bimanual reaching and example, grasping holding times were shortened. An adjustable table
grasping tasks tailored to individual motor capabilities, with was integrated to enhance comfort and accessibility for patients of
dynamically adjustable difficulty. The aesthetics of the platform different sizes and mobility levels. Additionally, control algorithms
were carefully chosen to avoid an uncanny-valley effect, virtual were refined to ensure smoother transitions between mirrored and
objects and hand representations are stylized rather than independent movements. These updates optimized the system for
hyperrealistic, while still providing an immersive experience. A usability and adaptability across diverse rehabilitation needs.
graphical user interface (GUI) for therapists was found to be
essential for enabling real-time customization of parameters and 2.3.6 P5 – Second software release
monitoring of patient progress. Notably, therapists can visualize The second software release resolved the issues identified in
patients’ actions in VR through a dedicated monitor, allowing them Phase P3 and incorporated updates based on the refined system
to observe kinematics movements and EMG-driven gestures live as requirements from Phase P4. Key improvements included a
part of the feedback mechanism. An initial GUI design underwent a configuration file that enabled therapists to customize
30-min click dummy evaluation followed by therapist feedback, experimental parameters prior to sessions, such as spatial
ensuring alignment with patients’ needs. This phase produced the tolerance for successful target reach, maximum trial duration
initial system architecture, outlining the integration of hardware (timeout), and trial success time (holding time required to
components, software modules, and control algorithms, providing successfully reach and perform a gesture). Visual and auditory
the basis for Phase P2. feedback mechanisms were enhanced for better clarity and
responsiveness. Additionally, adjustments to target orientation
2.3.3 P2 – First software release and placement were made to address positioning issues, such as
The first software version was developed based on the interactions with the table or targets placed in physiologically
specifications defined in Phase P1. Key features included challenging positions, ensuring better accessibility without
calibration tools, movement control strategies, task design, and compromising therapeutic effectiveness. These updates marked a
real-time feedback. The calibration tools ensured alignment significant advancement, addressing limitations from earlier phases
between the virtual and physical setups, enabling precise and preparing the system for evaluation with stroke patients.
interaction with the virtual environment and tailoring
experimental parameters to the patient’s motor capabilities. The 2.3.7 P6 – Evaluation of the system by therapists
movement control strategies allowed the virtual more affected limb and stroke patients
(VMAL) to be driven by the patient’s residual motor capabilities, The revised software was tested with three chronic stroke
with dynamic adjustable parameters to mirror the movements of the patients during a pre-pilot assessment to evaluate usability and
real less affected limb (RLAL) to amplify the movements of the real gather additional feedback. Patients with varying levels of
more affected limb (RMAL). The task design incorporated bimanual impairment participated in multiple sessions (Table 2), during
reaching and grasping exercises, carefully designed to offer engaging which therapists dynamically adjusted parameters to tailor tasks
and challenging experiences while promoting motor recovery. Real- to each patient’s individual capabilities. Observations and feedback
time feedback mechanisms delivered visual and auditory cues to from both therapists and patients were instrumental in informing
guide task execution and enhance patient engagement. This version further refinements in the subsequent development phase.
served as a proof of concept, demonstrating the feasibility of
integrating virtual reality, motion capture, and myoelectric 2.3.8 P7 – Second refinement of the system
control into a cohesive rehabilitation platform. requirements and specifications
Based on the feedback gathered during Phase P6, several
2.3.4 P3 – Initial evaluation of the system by refinements were identified as required to enhance the system’s
healthy participants and therapists usability and clinical effectiveness. These included updating the
Healthy participants tested the system under therapist virtual hand design, optimizing the virtual environment, refining
supervision to identify usability and functionality issues. The target placement algorithms, and addressing issues related to

Frontiers in Bioengineering and Biotechnology 06 frontiersin.org

De Pasquale et al. 10.3389/fbioe.2025.1655416

TABLE 2 Participants included in the pre-pilot usability assessment. Demographic and clinical data of the participants to the pre-pilot assessment.

Patient Age Gender Stroke type More affected limb Time since stroke (years) Session N° FMA score
1 33 Male Hemorrhagic Right 0.4 4 56

2 67 Female Ischemic Left 18.92 3 18

3 41 Female Ischemic Left 4.47 1 36

compensatory movement strategies. Updating the virtual hand to or during the study. The study was conducted in accordance with
design involved enhancing its color and appearance to improve the Declaration of Helsinki and approved by the ethical review board
patient embodiment and reduce confusion during tasks. Optimizing of FSL (Prot. CE/PROG.790). Written informed consent was
the virtual environment focused on removing non-interactable obtained from all participants before the experimental
objects from the virtual environment to minimize distractions sessions began.
and improve task focus. Refining target placement algorithms The pilot study lasted 1 month, with each participant completing
aimed to improve accessibility, particularly for lateral targets, rehabilitation sessions three times per week. Tasks involved
while maintaining their therapeutic value. To address the issues bimanual reaching and grasping, calibrated to each patient’s
identified with the presence of compensatory movements it was individual movement range. Real-time feedback was provided
decided to optimize the control algorithms and to provide therapists during the tasks to guide performance, while therapists
with tools to better manage these behaviors. Additionally, the need dynamically adjusted assistance parameters and task difficulty to
to develop a comprehensive operation manual to assist therapists in ensure the optimal balance between challenge and feasibility. The
system calibration, parameter adjustments, and managing primary focus was to evaluate the system’s ability to support motor
compensatory movement strategies effectively was identified. recovery and maintain patient engagement throughout the
These refinements were deemed critical to delivering a more rehabilitation process. Validated questionnaires were used to
effective, intuitive, and engaging rehabilitation platform. assess the system’s usability, feasibility and patient experience.
For the evaluation of the usability, the feasibility of the
2.3.9 P8 – Final software release developed rehabilitation system and patient experience, the
The final version of the system successfully addressed all the following questionnaires were administered to the pilot
remaining usability issues, incorporating feedback and experiment participants: the User Satisfaction Evaluation
improvements to enhance its effectiveness. Key refinements Questionnaire (USEQ, (Gil-Gómez et al., 2017)), for evaluation
included embodiment improvements, with optimized virtual of the evaluation of the VR system usability with range from 6 to
hand designs for better interaction and engagement, and 30; the Visual Analogical Scale (VAS) with range from 0 to 10 with
therapist interface enhancements, featuring improved GUI respect to subjective motivation and satisfaction related to exercise;
functionality to allow seamless real-time adjustments of the Pittsburgh Participation to rehabilitation Scale (PPRS, (Iosa
parameters. Calibration refinements ensured consistent accuracy et al., 2021)) compiled by the researcher/therapist to report the
through improved alignment procedures. patient’s participation levels in the exercise on a Likert scale ranging
While trunk compensation remained a challenge, therapists from 1 to 6 and the National Aeronautics and Space Administration
were equipped with tools and guidelines to manage these Task Load Index (NASA-TLX, (Hart and Staveland, 1988)) for the
behaviors effectively. A comprehensive user manual was also multidimensional subjective assessment that rates perceived
provided to therapists prior to the pilot experiment. The manual workload to assess a task, system, or team’s effectiveness or other
detailed the hardware and software components of the system, aspects of performance.
described operating procedures, outlined the initial calibration The Fugl-Meyer Assessment (FMA) for the upper limb was used
process, and explained policies for adjusting assistance to evaluate motor recovery and functional ability. This evidence-
parameters during training. This version was deemed ready for based scale (Deakin et al., 2003) assesses motor recovery across
clinical evaluation, incorporating critical updates and support multiple stages, with individual items scored on an ordinal scale
materials to facilitate effective use and study of the system in from 0 (unable to complete the task) to 2 (successfully completed).
real-world scenarios. The FMA provided a standardized measure of the system’s
effectiveness in promoting motor improvements.
2.3.10 P9 – Evaluation of the system with a Instrumental data were recorded through the VVITA system’s
pilot study integrated sensors. The recorded kinematic from the hands (3D
Phase P9 involved a pilot study designed to evaluate the position and rotation the back of the hand) was tracked at 30 Hz,
feasibility, usability, and clinical impact of the final VVITA filtered using a fourth-order Butterworth low-pass filter with a 3 Hz
system. Four stroke patients were recruited for this phase, cutoff frequency, and analyzed so to extract assessment variables
including two in the chronic phase (at least 1 year post-stroke) such as performance scores, assistance parameters, maximum speed,
and two in the subacute phase (less than 1 year post-stroke) and range of motion (ROM).
(See Table 3). Both clinical and instrumental evaluations were conducted at
To avoid confounding factors, participants had not received the beginning (T0) and end (T1) of the rehabilitation protocol to
additional treatments, such as botulinum toxin, in the 2 weeks prior assess the system’s impact. Feedback from patients and therapists

Frontiers in Bioengineering and Biotechnology 07 frontiersin.org

De Pasquale et al. 10.3389/fbioe.2025.1655416

TABLE 3 Participants included in the pilot study. Demographic and clinical data of the participants to the pilot assessment.

Patient Age Gender Stroke More affected Dominant Time since stroke Session FMA
type limb hand (months) N° score
E1 39 Female Ischemic Left Right 48.7 13 38

E2 65 Male Ischemic Right Right 284.3 12 24

E3 72 Male Ischemic Left Right 0.8 4 16

E4 77 Female Ischemic Right Right 0.5 2 18

was also collected after each session to identify any remaining

usability challenges and gauge overall satisfaction with the
rehabilitation experience.

2.4 Statistical analysis

We evaluated the variation in overall assistance level, Fugl-

Meyer index, and maximum speed across sessions (Se) using a linear
mixed model (LMM). This model accounted for interindividual
variability by including participants as a random effect. The
experimental factor (Se) was treated as a fixed effect with
categorical (dummy) variables. Data were fitted with the model
in Equation 1:
Y
u0 + α0 Se + ϵ (1)

where u0 is the individual intercept and accounts for inter-individual

FIGURE 2
differences, α0 is the fixed-effect slope, thus the modulation of the VVITA Setup. (A) Virtual reality headset Vive by HTC. (B) EMG
response variable by the factor Se. As data represent a continuous bracelet Myo by Thalmic Labs. (C) HTC tracker. (D) HTC controller. (E)
Laptop. (F) One of the two HTC Vive base stations. (G) Adjustable table.
variable, they were fitted with a LMM (Matlab, function fitlme). (H) Therapist. (I) Physiotherapist.
Estimation of model parameters were based on the maximum
likelihood using Laplace approximation.

parameter (α) and a capability parameter (β) for both the distal and
3 Results proximal components of the more affected limb. The outcomes of
each development phase, as well as the results of the final evaluation
A UCD approach was employed to develop a novel system for of the system with a pilot study are reported in the following two
upper-limb post-stroke motor rehabilitation by modifying a sections. For the pilot study, metrics related to system usability, user
system initially created to treat phantom limb pain in experience and task performance are reported.
amputees. The training activities, assistive algorithms and user
interface were developed and refined through 65 interactions
among all the members of the development team, achieving a 3.1 Outcomes of incremental development
synthesis between technological innovation and rehabilitation
principles. 3.1.1 VVITA system
Following an initial analysis of the patients’ needs and The VVITA system is based on the VITA system with which it
rehabilitation goals, a bimanual reaching task was selected as the shares the hardware architecture. Differently from the VITA system,
primary training activity. Virtual mirroring (Saleh et al., 2017; in VVITA both trackers are placed on the dorsum of the patient’s
Giroux et al., 2018; Mekbib et al., 2021; Hsu et al., 2022; da Silva hands (Figure 2). Several software solutions were developed and
Jaques et al., 2023), using both kinematic signals for hand position adopted to provide a virtual mirror therapy assisted by the hand
and EMG signals for hand gesture, was defined as the key assistive kinematic and gesture prediction of the less affected arm depending
approach to enhance functional recovery. Initially, 3 different on the level of impairment and the functional restoration achieved
control algorithms were considered to provide virtual mirroring during the treatment. The system allows a patient to practice
of the more affected hand. These algorithms were implemented for rehabilitation exercises that simulate the performance of ADL in
pilot testing, each affecting the control modality and the guidance of an immersive VR environment, providing real-time visual feedback
the VMAL differently, based on the RLAL and the RMAL. of a bimanual reaching and grasping task in which the movement of
The amount of mirrored assistance provided to the VMAL, and a virtual limb reproduces and improves the movement of the paretic
the task difficulty were controlled by two parameters: an agency limb. The movement of the VMAL is displayed based on the

Frontiers in Bioengineering and Biotechnology 08 frontiersin.org

De Pasquale et al. 10.3389/fbioe.2025.1655416

FIGURE 3
Proximal assistance. (A) The agency parameter α modulates the contribution of the RLAL and RMAL to the motion of the VMAL. (B) The capability
parameter β tailors the task based on the patient’s residual motor abilities adjusting target placement distance. (C) A graphical representation of the
relationship between α and β: higher values of α and β denote more difficult tasks due to reduced assistance and/or larger workspace.

movement and the EMG activity of both the RMAL and the RLAL contribution of the RLAL and RMAL movements to the VMAL
recorded with the system integrated motion capture sensors and the movement. This parameter can be adjusted independently for
armbands as in the original VITA system. proximal (αp ) and distal (αd ) control. For proximal control, the
VMAL position in 3D space is calculated as a weighted combination
3.1.2 Assistive and control algorithms of RMAL and mirrored RLAL movements, defined as:
The VVITA system introduces innovative assistive and control
xVMAL
αp xRMAL +
1 − αp F[xRLAL ] (2)
algorithms that enable virtual mirror therapy, a functionality not
available in the original VITA system. The system supports In Equation 2, x represents the pose (i.e., position and
bimanual rehabilitation tasks by leveraging VR capabilities to orientation) of the hand in Cartesian coordinates, and F[x] is a
provide mirrored assistance for both limb movements (proximal mirroring function that can be defined according to different
control) and hand gestures (distal control). Proximal control utilizes methods (see below).
kinematic data recorded by motion trackers, while distal control uses For distal control, the VMAL hand gesture is determined by the
EMG data acquired from the Myo armband. EMG activity of the RMAL and RLAL:
To achieve precise control of the VMAL, the system employs
an agency parameter (α) (Figure 3A), which determines the qVMAL
αd qRMAL + (1 − αd )qRLAL (3)

Frontiers in Bioengineering and Biotechnology 09 frontiersin.org

De Pasquale et al. 10.3389/fbioe.2025.1655416

In Equation 3, q denotes a parameter controlling the degree of is randomly selected by the software each time it is launched. A table
hand opening and closing, representing the gesture of the hand is positioned in front of the patient, and virtual representations of
(i.e., the set of finger joint angles describing the configuration of the the patient’s hands, along with objects to be reached and grasped
fingers relative to the palm), as estimated from EMG signals. bimanually, are displayed. Two types of objects can be presented as
As αp ≈ 1 or αd ≈ 1, the VMAL replicates the RMAL’s targets in the virtual environment: a concertina (Figure 4A) and a
movement or gesture, providing complete agency and no ball (Figure 4B). All reaching and grasping tasks are designed to be
assistance. Conversely, as αp ≈ 0 or αd ≈ 0, the VMAL reflects symmetrical with respect to the target object, to ensure the
the RLAL’s motion or gesture, resembling conventional mirror effectiveness of the RLAL assistive guidance for the VMAL.
therapy. This flexibility allows the therapist to customize the
assistance levels according to the patient’s motor capabilities and 3.1.3.2 GUI: setup of experimental parameters and
the success rate of reaching tasks, choosing within ten predefined rehabilitation protocol
assistance levels ranging from zero assistance to full assistance. To meet the therapists’ specifications and ensure a user
The mirroring function F in Equation 2 transforms spatial experience aligned with their expectations, a dedicated GUI
positions to compute a reference position that helps guide the (Figure 4C) was developed to control experimental parameters
movement of the VMAL based on the motion of the RLAL. and guide the execution of the rehabilitation protocol. First, the
Specifically, the VMAL hand is displayed at a weighted average virtual environment is calibrated to the physical workspace by using
between the RMAL hand position and the mirrored reference HTC Vive controllers. During this step, the operator marks the
position. Three different methods were implemented to corners of the real table to ensure precise alignment of the
determine such reference position by transforming the position corresponding virtual table. Once this calibration is completed,
of the RLAL: stiff coupling, rubber band, and reference the Myo armbands and the tracker, worn by the patient, are
trajectories. In the stiff coupling method, the reference position is activated and recognized by the system, enabling it to distinguish
computed rigidly as the RLAL hand position mirrored across a between the less affected and more affected limbs. The patient then
vertical plane. The rubber band method mirrors the RLAL’s performs a series of calibrations tasks under the supervision of the
movement across the mid-sagittal plane, mimicking classical operator, aimed at defining motor capabilities and tailoring the
mirror therapy. The reference trajectories method maps the experimental parameters accordingly. First, a resting pose
reference position along predefined paths derived from the pre- calibration is conducted to determine the baseline hand position.
recorded RLAL trajectories, providing a structured reference This is followed by a proximal calibration to establish maximum
trajectory for the more affected limb. These algorithms ensure range of motion. Finally, during the distal calibration, EMG signals
adaptable, task-specific assistance, facilitating rehabilitation are recorded to train the gesture recognition model. After these
tailored to each patient’s motor recovery. preliminary steps, the operator selects the patient’s profile and
In addition to agency parameters, the system incorporates customizes the rehabilitation approach by choosing one of three
capability parameters (β) to adjust task difficulty dynamically assistance modalities: stiff coupling, rubber band, or trajectories.
(Figure 3B). The proximal capability parameter (βp ) regulates Additionally, the software allows fine-tuning of four key parameters
target placement, with levels ranging from 0 (targets placed (αp, αd, βp, βd) to adjust task difficulty, provides performance metrics
within the RMAL’s reachable range) to 1 (targets placed at the such as the number of successful trials, and securely stores patient-
RLAL’s maximum range). Similarly, the distal capability parameter related data.
(βd ) adjusts the tolerance for successful hand gestures, requiring
minimal muscle activation if βd ≈ 0 and close-to-maximal effort if 3.1.4 Rehabilitation protocol and task design
βd ≈ 1. Patients performed rehabilitation exercises with the VVITA
The relationship between the assistance parameters and task system over the course of 1 month, with sessions scheduled three
difficulty is defined such that higher parameter values corresponded times per week, lasting 30 min each. During each session, the patient
to reduced assistance levels, thereby increasing task difficulty. performs a series of movements grouped into blocks, with each
Specifically, as α or β increase, patients are required to rely more block consisting of 12 bimanual reaching and grasping movements.
on their own motor abilities to complete tasks, promoting active The target to be reached is randomly selected from the
engagement and rehabilitation. Conversely, lower parameter values predefined set of locations (Figure 5). Target positions vary in
provided greater assistance, adequate for more severe impairments. height and lateral deviation relative to the initial recorded resting
Thus, this framework allows the therapist to dynamically adjust the position of the hands on the table. Positions are defined as
difficulty of the task in response to the patient’s changes in agency xtar
xtar [r, θ, h, βp ], where r is the radial distance from the
and capabilities (Figure 3C). starting hand position to the target center, θ is the target
azimuth, and h is the target height. The distance parameter r is
3.1.3 Virtual environments and GUI dynamically adjusted as a fraction of the maximum distance
3.1.3.1 Virtual environments recorded during calibration (r
βp rmax ). The azimuth θ and
The VVITA software facilitates bimanual upper-limb reaching height h of the targets are configured to span the entire
tasks in an immersive VR environment. The virtual environment calibration space, ensuring comprehensive coverage of the
consists of three distinct scenarios: a house, a garden near the house, patient’s reachable workspace. Similarly, the muscle activation
and a green space by a pond, designed to provide engaging and level required for successful hand gestures are determined by βd ,
varied settings for rehabilitation based on real life interactions to with higher values requiring greater activation intensity to achieve a
reduce the cognitive load of learning new mechanisms. One scenario successful grasp.

Frontiers in Bioengineering and Biotechnology 10 frontiersin.org

De Pasquale et al. 10.3389/fbioe.2025.1655416

FIGURE 4
Virtual environment and GUI. The virtual environment consists of a room in a house with a desk and two different objects, a concertina (A) and a
beach ball (B). Participants are instructed to reach for these objects with both hands and grab them using the specified gesture—fingers closed for the
concertina and fingers open for the beach ball. When the correct position and gesture are maintained for a set duration, the object progressively turns
green. (C) Shows the sequential workflow of the therapist’s GUI.

FIGURE 5
Target placement. Example target placements for βp
1. (A) 3D isometric view; (B) top view; (C) lateral view; (D) frontal view of the 12 targets (green
spheres), the head (black sphere) and RLAL (blue sphere) and RMAL (red sphere) resting poses.

Frontiers in Bioengineering and Biotechnology 11 frontiersin.org

De Pasquale et al. 10.3389/fbioe.2025.1655416

Each trial begins with the patient’s hands resting comfortably on unreachable targets. This was solved by enhancing the alignment
the table, close to the body, with elbows flexed at 90°. Midway procedure to ensure accurate calibration. Additionally, proximal
through the session, the patients remove the head-mounted display assistance parameters sometimes led to unnatural movements in
and take a 3-min break. Target positions for each block are “stiff coupling” and “rubber band” modes; the “stiff coupling”
randomly selected (without replacement) from the 12 predefined method was refined as the default due to its closer resemblance to
positions. At the beginning of each session, the patients calibrate natural kinematics. Moreover, patients reported visual distractions
their range of movement by performing maximal motions to the from non-interactable objects and flickering in the virtual
boundaries of their workspace with both the less-affected and most- environment, prompting the removal of unnecessary objects and
affected limbs. The maximum height and forward reach recorded the introduction of a more neutral virtual hand color to improve
during calibration are used to set the target placement distances for embodiment. Calibration procedures also required adjustments, as
all subsequent trials. fixed times led to inaccurate motion range estimates, and extended
The two target objects, the concertina and ball, are alternated trial durations caused fatigue. These were resolved by introducing
throughout the session to require different hand gestures for each operator-controlled calibration and reducing trial durations from
target. The concertina requires a fist gesture (fingers closed), while 30 to 20 s. Patients with limited functionality often compensated
the ball requires a hand extension gesture (fingers opened). Both for impaired limb movements with excessive trunk motion, which
objects are equipped with two handles positioned on opposite sides, therapists mitigated through physical interventions rather than more
guiding the required hand placement for successful interaction. A invasive solutions suggested like belts or additional tracking systems.
reaching movement is considered successful when both virtual Target placement issues were also addressed by introducing
hands are within 0.05 m of the target handles. configurable target positions and sequences, along with a threshold
Real-time feedback on hand placement and gesture success is to prevent unreachable target positions. These iterative improvements
provided within the VR environment. When the virtual hands are highlight the significance of integrating patient and therapist feedback
within the required tolerance, the handles’ color changes from in the system refinement, ultimately enhancing its usability,
orange to green. Additionally, a visual indicator of the required adaptability, and effectiveness in post-stroke rehabilitation.
hand gesture disappears once the gesture is correctly performed.
Upon meeting both position and gesture requirements, the object
progressively turns green, visually signaling the required holding 3.2 Outcomes of the pilot study
time (Figure 4A). Two types of auditory feedback are provided: a
positive cue for successful completion of the task and a negative cue In this section, clinical and instrumental results from four stroke
if the trial exceeds a 20-s time limit. The trial duration is limited to patients recorded during the pilot study are reported. Outcomes
20 s to prevent fatigue and to maintain task efficiency. from two chronic patients (E1 and E2) who performed 13 and
12 sessions respectively (including the first familiarization session)
3.1.5 Task adaptation and performance monitoring are presented in more detail. For the two subacute patients (E3 and
At the end of each block, the therapist adjusts the proximal and E4) who performed 4 and 2 sessions respectively (including the first
distal agency (αp , αd ) and capability (βp , βd ) parameters based on familiarization session), only feasibility and acceptability
the patient’s performance score π j
π(t1 , . . . , tTj ), calculated for questionnaire responses are reported.
block j of Tj trials. The performance score determines task difficulty
adjustment: 3.2.1 Performance assessment
During each session, assistance parameters (α and β, proximal
• >90% Success Rate (≥11 successful trials): Task difficulty and distal) were adjusted at the end of each block according to the
is increased. performance in that block, as described above (Figure 6). In Figure 7,
• 70%–90% Success Rate (9–10 successful trials): Task difficulty the assistance parameters and the performance for each block across
remains unchanged. all the sessions are reported for patient E1 (A), who had a higher
• <70% Success Rate (≤8 successful trials): Task difficulty level of residual functionality and patient E2 (B), who had a lower
is decreased. level of residual functionality. As shown in the performance graphs,
when performance was below 70% or above 90%, the difficulty was
This adaptive framework ensures that the task remains within an decreased or increased by adjusting the assistance parameters
optimal challenge range, maintaining a success rate of 70%–90% to accordingly. Thus, the therapist was successful in keeping the
promote engagement and motor improvement throughout the level of performance around 80%. During the first session, a
rehabilitation process (Figure 6). The selection of the parameter familiarization phase was performed in which the maximum level
to be adjusted, either to increase or decrease task difficulty based on of α proximal and then β were set to let participants explore the
the score, was made by the therapist, guided by clinical expertise and software capabilities. A medium level of difficulty, based on
real-time observation of the patient’s needs. participant performances and therapists’ observations, was then
selected to start the therapy.
3.1.6 Challenges, limitations and solutions Figure 8 shows the distributions of parameters used during the
During the development of the VVITA system, several challenges rehabilitation protocol for E1 (A) and E2 (B). For each patient, the
emerged, which were systematically addressed following the UCD distribution of the proximal (left column) and distal (right column)
approach. One key issue was the misalignment between the virtual assistance parameters are shown. The figure highlights a clear
and real tables, causing visuo-proprioceptive mismatches and difference between E1 and E2. For E1, the proximal parameters

Frontiers in Bioengineering and Biotechnology 12 frontiersin.org

De Pasquale et al. 10.3389/fbioe.2025.1655416

FIGURE 6
Assistance and difficulty parameters adjustment procedure. Decision block diagram illustrating the process for adjusting difficulty parameters based
on the success performance of the previous reaching block (12 trials), as utilized by the operator.

are both close to one (mean ± standard deviation (SD): confirming that the therapist correctly implemented the policy for
αp
0.91 ± 0.10; βp
0.75 ± 0.11), indicating less assistance. In maintaining a constant level of task difficulty.
contrast, for E2, who was more impaired than E1, both proximal Participants E1 and E2 were evaluated clinically at the beginning
parameters were lower than 0.6 (mean ± SD: αp
0.39 ± 0.01, (T0) of the rehabilitation protocol and at the end (T1) using the
βp
0.23 ± 0.05). Furthermore, for E1 the parameters overtime Fugl-Meyer assessment scale, which has 3 points (index: 0, 1, 2) for
became closer to 1 than for E2.’ with ‘Furthermore, for E1 the each item of the upper-limb motor function assessment.’ with
parameters, over time, became closer to 1 than for E2. This indicates ‘Participants E1 and E2 underwent clinical evaluation at T0 and
that, while maintaining the same performance level, E1 was able to T1 using the Fugl-Meyer assessment scale, which employs a 3-point
perform the task with greater difficulty compared to E2. However, scale (0, 1, 2) for each item assessing upper-limb motor function.
since the therapist could modulate task difficulty by adjusting both Figure 9C shows the evolution of the Fugl-Meyer motor function
proximal and distal assistance parameters, we also computed an assessment index for each patient (different colors) for each
overall assistance index to summarize tasks difficulty. evaluation (T0, T1), expressed as percentage of the maximum
αp +βp +αd +βd
The overall assistance level 1 − 4 decreased for both score (66 points). For E1 and E2 the motor function upper-limb
patients between T0 and T1 (Figure 9A, different colors). Assistance score increased during the rehabilitation protocol from 38 to 39
data of each patient for each block were fitted with the LMM model (57.6%–59.1% of the maximum score), corresponding to an increase
of Equation 1 (R2
0.72), which revealed a significant main effect of of 1.52% for E1 and from 24 to 28 (36.4%–42.4%), corresponding to
session (p < 0.001). an increase of 6.06%, for E2. However, when the Fugl-Meyer
The therapist was instructed to change the assistance parameters assessment index data were fitted with the LMM model of Eq. 1
in order to maintain performance at around 80% in each block. (R2
0.99), the main effect of session was not significant (p
0.14).
Figure 9B shows the overall performance for each patient. Maximum speed was estimated for RMAL and RLAL in each
Performance was 87 ± 9 (mean ± SD) for E1 and 84 ± 8 for E2, trial to evaluate its evolution during the rehabilitation protocol. The

Frontiers in Bioengineering and Biotechnology 13 frontiersin.org

De Pasquale et al. 10.3389/fbioe.2025.1655416

FIGURE 7
Block parameter selection and performance. Each panel shows the four assistance parameters (αp , βp , αd , βd ) used for each block and the
participants’ performance (number of successful trials over the number of trials) for patient E1 (A) and E2 (B).

average maximum speed values for each limb for all patients are during the maximum range of motion calibration at the beginning of
shown in Figure 10A for T0 and T1. Maximum speed averaged the T0 and T1 sessions for both E1 (panel A) and E2 (panel B). Left-
within each session were fitted with the LMM model of Equation 1. hand movements are displayed on the left side of the figure, while
(RMAL R2
0.92, RLAL R2
0.88), which revealed a significant right-hand movements are shown on the right. RMAL trajectories
main effect of session for both limbs (RMAL p
0.047, are represented in red for T1 and in a lighter shade (orange) for T0,
RLAL p < 0.001). whereas RLAL trajectories are shown in green for T1 and in a lighter
Finally, to quantify the range of movement, an estimation of the shade (yellow-green) for T0.
volume of the convex hull of the trajectories recorded during the
calibration procedure, performed at the beginning of each session, 3.2.2 Usability assessment
was obtained (Matlab, convhull function). Figure 10B shows the All participants reported a good level of usability with the USEQ
average RMAL (A) and RLAL (B) convex hull volume values for scale, even when motivation and perceived satisfaction were low as
each hand path for the two patients for the first session T0 and the in the case of patient E3; further confirmed by the researcher’s
last session T1. Convex hull volume data were fitted with the LMM evaluation through the administration of the PPRS scale
model in Equation 1 (RMAL R2
0.97, RLAL R2
0.09), which (Figure 12A). Usability assessed by patients every session with
revealed a non-significant main effect of sessions for both limbs NASA-TLX (Figure 12B), reported high levels of cognitive and
(RMAL p
0.34, RLAL p
0.59). physical demand; good levels related to the time domain
To highlight the differences in hand trajectories between the first (NASA_T) which indicates a general acceptability of the training,
and last sessions, Figure 11 illustrates the spatial paths recorded while, there are differences between the two chronic (E1 and E2) and

Frontiers in Bioengineering and Biotechnology 14 frontiersin.org

De Pasquale et al. 10.3389/fbioe.2025.1655416

FIGURE 8
Parameters distribution. Figure shows the distribution of proximal (left column) and distal (right column) assistance parameters for E1 (A) and E2 (B).
For each time interval Bin (3 sessions), the mean value (ellipse center) and covariance (95% CI. ellipse) of α and β are reported. Color saturation represents
the temporal evolution of the distribution.

subacute (E3 and E4) patients with respect to frustration (NASA_F) Traditional rehabilitation approaches, such as mirror therapy
and self-assessment related to perceived success associated with the and conventional VR systems, have demonstrated potential but are
training (NASA_P). constrained by limited adaptability and personalization. Mirror
therapy primarily relies on visual feedback to facilitate neural
reorganization but cannot dynamically adjust to individual
4 Discussion patient needs (Altschuler et al., 1999; Thieme et al., 2012).
Similarly, conventional VR systems provide engaging
The VVITA system represents a significant advancement in environments but lack advanced control strategies to integrate
stroke rehabilitation approaches based on virtual reality, patient-specific motor capabilities effectively (Bohil et al., 2011;
myoelectric control, and exergames, developed using a rigorous Laver et al., 2017). VVITA addresses these limitations by
UCD methodology. This approach actively involved patients, combining immersive VR environments with dynamic task
therapists, and clinicians, ensuring the system’s features modulation based on real-time kinematic and EMG data, in
directly address real-world rehabilitation challenges. The which activity from the less and more affected arm are combined
system’s integration of immersive VR, adaptive assistance to control the virtual representation of the affected arm with
algorithms, and EMG control sets it apart from traditional and adaptive degree of relative contribution. This approach
earlier VR-based rehabilitation approaches. This discussion overcomes the “learned non-use” phenomenon (Ballester et al.,
provides a critical analysis of the VVITA system’s outcomes, its 2015) by enabling adaptive assistance tailored to the patient’s
development process, and its position in the landscape of stroke motor capabilities. Additionally, the use of EMG-driven
rehabilitation technologies. interfaces promotes targeted motor recovery by integrating

Frontiers in Bioengineering and Biotechnology 15 frontiersin.org

De Pasquale et al. 10.3389/fbioe.2025.1655416

FIGURE 9
Training assistance, performance and clinical outcomes. The figure presents the assistance level, the overall performance and the clinical evaluation
FM index for each patient (different colors). (A) Mean and SD of the assistance level provided to each patient at T0 and T1. (B) Overall performance,
expressed as the percentage of successful trials over total trials for each patient. (C) Clinical evaluation of upper-limb motor function at T0 and T1,
normalized as a percentage (with 66 corresponding to 100%), for each patient. Statistical significance between sessions is reported as *** for
p < 0.001, ** for p < 0.01, and * for p < 0.05.

FIGURE 10
Maximum Speed and Movement Range. The figure presents the maximum speed and the volume of the maximum movement range achieved by
each patient (different colors), evaluated during the rehabilitation protocol and the initial calibration procedure at T0 and T1 sessions. (A) The left panel
shows the maximum speed for the RMAL, while the right panel shows the maximum speed for the RLAL. (B) The left panel displays the convex hull
estimated for RMAL hand paths, while the right panel shows the convex hull for RLAL hand paths, both assessed during the initial calibration of
maximum range movements.

muscle activity into rehabilitation tasks (Song et al., 2013). These VVITA incorporated gamified environments to sustain patient
features position VVITA as a more engaging and effective motivation (Burke et al., 2009). The UCD methodology ensured
alternative to traditional and earlier VR systems. that every aspect of the system, from its adaptive algorithms to its
The VVITA system’s design and development were informed by user interface, was optimized for therapeutic effectiveness and
65 iterative interactions with clinical and technical stakeholders. patient usability (Meyer et al., 2019; Ríos-Hernández et al., 2021).
These interactions highlighted key shortcomings in existing Comparatively, VVITA’s UCD methodology builds on established
rehabilitation systems, such as insufficient adaptability and a lack practices in rehabilitation technology design. For instance, it has
of engaging, task-specific scenarios. Informed by these insights, been demonstrated that the iterative, UCD approach highlights its

Frontiers in Bioengineering and Biotechnology 16 frontiersin.org

De Pasquale et al. 10.3389/fbioe.2025.1655416

FIGURE 11
Hand trajectories during maximum range of motion calibration. The figure shows the spatial paths of the hand during the maximum range of motion
calibration recorded at the beginning of the T0 and T1 sessions for E1 (panel (A)) and E2 (panel (B)), with both RMAL and RLAL. The left side of the figure
represents left-hand movements, while the right side represents right-hand movements. RMAL trajectories are shown in red for T1 and in a lighter shade
(orange) for T0. Similarly, RLAL trajectories are shown in green for T1 and in a lighter shade (yellow-green) for T0.

value in developing effective lower-limb rehabilitation systems do not have any type of control and adaptation to the paretic limb
(Laffranchi et al., 2021; Semprini et al., 2022). Similarly, haptic limiting the potentiality of the VR-assisted arm sensory motor
robot-based telerehabilitation systems emphasize the importance of rehabilitation. An important novelty of the present work is the
user-friendly interfaces and personalized therapy (Ivanova et al., myoelectric control of the paretic limb in virtual reality. This has
2017). VVITA extends these principles by integrating advanced two direct clinical positive consequences: i) it enhances the
EMG control and real-time task modulation, offering a novel effectiveness of rehabilitation thanks to the greater integration
approach to stroke rehabilitation. of motor sensors consistent with the task; ii) it makes virtual reality
Preliminary testing with four stroke patients, two chronic and exercise more adaptable to the patient’s motor function both
two sub-acute, provided valuable insights into VVITA’s feasibility intersession and with respect to the criteria for patient
and effectiveness. Chronic patients exhibited higher engagement inclusion, allowing in fact to expand the beneficiaries.
and notable improvements in motor performance, thanks to the Moreover, therapists benefit from a real-time monitoring and
benefits of tailored interventions (Torrisi et al., 2021; De Luca et al., feedback loop introduced by the system, which displays live
2024; De Pasquale et al., 2024). Instrumental assessments revealed kinematic, EMG, and performance metrics, enabling immediate
significant improvements in maximum movement speed, adjustments to assistance levels and task parameters as therapy
correlating with clinical outcomes measured by the Fugl-Meyer progresses.
Assessment. This correlation underscores the system’s potential The novelty of the VVITA approach lies in its integration of VR,
to integrate objective metrics with subjective clinical evaluations, EMG control, and adaptive assistance, features that distinguish it
enhancing the assessment of rehabilitation effectiveness. from existing rehabilitation systems. For example, traditional VR
Interestingly, the pilot study revealed differences between systems lacked the capability to dynamically adjust task difficulty
patients in compliance and outcomes, with chronic patients based on real-time performance data (Ballester et al., 2015;
responding more positively to the training protocol. This aligns Hoermann et al., 2017). Similarly, other systems focused on
with prior research which indicates that gamified VR interventions immersive environments but did not integrate advanced control
are particularly effective in maintaining motivation and improving strategies to modulate assistance levels or enhance task specificity
motor function (Domínguez-Téllez et al., 2020). However, sub-acute (Lee et al., 2011; Laver et al., 2017; Ceradini et al., 2024; Mani
patients faced challenges such as mood-related issues and medical Bharathi et al., 2024).
complications, highlighting the need for further customization and The VVITA system’s ability to combine kinematic and EMG
support in this patient subgroup. data for personalized therapy represents significant improvement
Both immersive and non-immersive virtual reality systems are with respect to previous approaches presented in literature. Indeed,
now widespread in research and in clinical practice, however they the importance of adaptive systems in achieving meaningful motor

Frontiers in Bioengineering and Biotechnology 17 frontiersin.org

De Pasquale et al. 10.3389/fbioe.2025.1655416

Data availability statement

The raw data supporting the conclusions of this article will be
made available by the authors, without undue reservation.

Ethics statement
The studies involving humans were approved by Ethical review
board of IRCCS Santa Lucia Foundation (Prot. CE/PROG.790). The
studies were conducted in accordance with the local legislation and
institutional requirements. Written informed consent for
participation in this study was provided by the participants’ legal
guardians/next of kin.

Author contributions
PDP: Writing – original draft, Conceptualization,
Writing – review and editing, Investigation, Formal
Analysis, Project administration, Validation, Methodology,
FIGURE 12
Visualization, Data curation. DDB: Writing – review and
Feasibility and Usability scales. The figure shows the mean and SD editing, Investigation, Conceptualization, Data curation,
of Feasibility (A) and Usability (B) values. For the feasibility outcome, Validation, Project administration. MR: Methodology, Validation,
the scales represented in different colors (PPRS, VAS 742 M, VAS S, and
USEQ) are reported for the 4 patients. For usability, the 6 items of Investigation, Conceptualization, Writing – review and editing.
the NASA scale (C, cognitive demand; P, physical demand; T, temporal DJB: Methodology, Writing – review and editing, Investigation,
demand; S, perceived success; E, effort; F, frustration) are reported for Validation, Conceptualization. AM: Conceptualization,
the 4 patients, each represented in different colors.
Writing – review and editing, Methodology. DB:
Writing – review and editing, Methodology, Conceptualization.
EC: Writing – review and editing, Conceptualization,
recovery aligns with findings in the literature (Sadarangani et al., Methodology. DM: Methodology, Conceptualization,
2017; Morone et al., 2019). However, further studies are needed to Writing – review and editing. CN: Writing – review and editing,
compare VVITA directly with other advanced systems, such as Software, Methodology, Conceptualization. MN: Software,
robotic exoskeletons or hybrid VR systems, to assess its clinical Methodology, Conceptualization, Writing – review and editing.
efficacy and cost-effectiveness. EF: Writing – review and editing, Software, Conceptualization,
Nonetheless, the very small sample size (four patients) is Methodology. JSM: Conceptualization, Writing – review and
an important limitation of this study. Such sample size is editing, Methodology, Software. MRS: Software, Writing – review
appropriate for a pilot study assessing the feasibility of a novel and editing, Conceptualization, Methodology. CC:
approach but limits the generalizability of the findings. Therefore, Conceptualization, Writing – review and editing, Software,
the results should be interpreted with caution, and further Resources, Funding acquisition, Methodology, Validation. GM:
research with larger, more diverse cohorts is necessary to Methodology, Validation, Conceptualization, Supervision,
validate these preliminary observations. Building on the Funding acquisition, Writing – review and editing, Resources.
promising results of the pilot study, a randomized controlled AdA: Funding acquisition, Writing – review and editing,
trial will evaluate VVITA’s clinical efficacy compared to Resources, Validation, Supervision, Methodology,
conventional rehabilitation methods. Similarly to what is Conceptualization.
shown here, key outcomes will include clinical indices (e.g.,
Fugl-Meyer Assessment), kinematic performance metrics, and
patient-reported satisfaction. Funding
The author(s) declare that financial support was received for
5 Conclusion the research and/or publication of this article. This work was
partially supported by the German Federal Ministry of Education
VVITA provides a novel and adaptable approach to stroke and Research (BMBF) under the Robotics Institute Germany
rehabilitation, addressing critical gaps in traditional and VR- (RIG); by the project (VVITA) awarded to Deutsches Zentrum
based systems. By leveraging UCD principles, advanced control für Luft-und Raumfahrt e.V. (DLR, Germany) by Helmoltz
mechanisms, and engaging VR environments, the system holds Association e.V.; by the Next-Generation EU Project (NGEU)
significant potential to improve patient outcomes and set a new National Recovery and Resilience Plan (NRRP), project
standard for adaptive rehabilitation technologies. MNESYS (PE0000006) – A Multiscale integrated approach to

Frontiers in Bioengineering and Biotechnology 18 frontiersin.org

De Pasquale et al. 10.3389/fbioe.2025.1655416

the study of the nervous system in health and disease (DN. Generative AI statement
1553 11.10.2022) Spoke 1; and by Current Research Funds
2025, Ministry of Health, Italy. The author(s) declare that no Generative AI was used in the
creation of this manuscript.

Conflict of interest
Publisher’s note
Authors EF, JSM, and MRS were employed by NEEEU
Spaces GmbH. All claims expressed in this article are solely those of the authors
The remaining authors declare that the research was and do not necessarily represent those of their affiliated organizations,
conducted in the absence of any commercial or financial or those of the publisher, the editors and the reviewers. Any product
relationships that could be construed as a potential conflict that may be evaluated in this article, or claim that may be made by its
of interest. manufacturer, is not guaranteed or endorsed by the publisher.

References
Altschuler, E. L., Wisdom, S. B., Stone, L., Foster, C., Galasko, D., Llewellyn, D. M. E., DiCicco, M., Lucas, L., and Matsuoka, Y. (2004). “Comparison of control strategies for
et al. (1999). Rehabilitation of hemiparesis after stroke with a mirror. Lancet 353, an EMG controlled orthotic exoskeleton for the hand,” in IEEE International
2035–2036. doi:10.1016/S0140-6736(99)00920-4 Conference on Robotics and Automation, 2004. Proceedings. ICRA ’04. 2004, New
Orleans, LA, USA, 26 April 2004 - 01 May 2004 (IEEE), 1622–1627.
Ballester, B. R., Nirme, J., Duarte, E., Cuxart, A., Rodriguez, S., Verschure, P., et al.
(2015). The visual amplification of goal-oriented movements counteracts acquired non- Dohle, C., Püllen, J., Nakaten, A., Küst, J., Rietz, C., and Karbe, H. (2009). Mirror
use in hemiparetic stroke patients. J. NeuroEngineering Rehabil. 12, 50. doi:10.1186/ therapy promotes recovery from severe hemiparesis: a randomized controlled trial.
s12984-015-0039-z Neurorehabil Neural Repair 23, 209–217. doi:10.1177/1545968308324786
Bohil, C. J., Alicea, B., and Biocca, F. A. (2011). Virtual reality in neuroscience Domínguez-Téllez, P., Moral-Muñoz, J. A., Salazar, A., Casado-Fernández, E., and
research and therapy. Nat. Rev. Neurosci. 12, 752–762. doi:10.1038/nrn3122 Lucena-Antón, D. (2020). Game-based virtual reality interventions to improve upper
limb motor function and quality of life after stroke: systematic review and meta-analysis.
Burke, J. W., McNeill, M. D. J., Charles, D. K., Morrow, P. J., Crosbie, J. H., and
Games Health J. 9, 1–10. doi:10.1089/g4h.2019.0043
McDonough, S. M. (2009). Optimising engagement for stroke rehabilitation using
serious games. Vis. Comput. 25, 1085–1099. doi:10.1007/s00371-009-0387-4 Eng, K., Siekierka, E., Pyk, P., Chevrier, E., Hauser, Y., Cameirao, M., et al. (2007).
Interactive visuo-motor therapy system for stroke rehabilitation. Med. Bio Eng. Comput.
Ceradini, M., Losanno, E., Micera, S., Bandini, A., and Orlandi, S. (2024). Immersive
45, 901–907. doi:10.1007/s11517-007-0239-1
VR for upper-extremity rehabilitation in patients with neurological disorders: a scoping
review. J. NeuroEngineering Rehabilitation 21, 75. doi:10.1186/s12984-024-01367-0 Foley, N., Pereira, S., Salter, K., Meyer, M., Andrew McClure, J., and Teasell, R. (2012).
Are recommendations regarding inpatient therapy intensity following acute stroke
Chen, M.-H., Huang, L.-L., Lee, C.-F., Hsieh, C.-L., Lin, Y.-C., Liu, H., et al. (2015). A
really evidence-based? Top. Stroke Rehabilitation 19, 96–103. doi:10.1310/tsr1902-96
controlled pilot trial of two commercial video games for rehabilitation of arm function
after stroke. Clin. Rehabil. 29, 674–682. doi:10.1177/0269215514554115 Gil-Gómez, J.-A., Manzano-Hernández, P., Albiol-Pérez, S., Aula-Valero, C., Gil-
Gómez, H., and Lozano-Quilis, J.-A. (2017). USEQ: a short questionnaire for
Cheung, V. C. K., Piron, L., Agostini, M., Silvoni, S., Turolla, A., and Bizzi, E. (2009).
satisfaction evaluation of virtual rehabilitation systems. Sensors (Basel) 17, 1589.
Stability of muscle synergies for voluntary actions after cortical stroke in humans. Proc.
doi:10.3390/s17071589
Natl. Acad. Sci. U.S.A. 106, 19563–19568. doi:10.1073/pnas.0910114106
Giroux, M., Barra, J., Zrelli, I.-E., Barraud, P.-A., Cian, C., and Guerraz, M. (2018).
Choi, Y.-H., Ku, J., Lim, H., Kim, Y. H., and Paik, N.-J. (2016). Mobile game-based
The respective contributions of visual and proprioceptive afferents to the mirror illusion
virtual reality rehabilitation program for upper limb dysfunction after ischemic stroke.
in virtual reality. PLOS ONE 13, e0203086. doi:10.1371/journal.pone.0203086
RNN 34, 455–463. doi:10.3233/RNN-150626
Han, J., Lian, S., Guo, B., Li, X., and You, A. (2017). Active rehabilitation training
Cikajlo, I., Rudolf, M., Mainetti, R., and Borghese, N. A. (2020). Multi-exergames to
system for upper limb based on virtual reality. Adv. Mech. Eng. 9, 168781401774338.
set targets and supplement the intensified conventional balance training in patients with
doi:10.1177/1687814017743388
stroke: a randomized pilot trial. Front. Psychol. 11, 572. doi:10.3389/fpsyg.2020.00572
Hart, S. G., and Staveland, L. E. (1988). “Development of NASA-TLX (task load
Commissioner, O. (2020). FDA permits marketing of first game-based digital
index): results of empirical and theoretical research,” in Advances in psychology. Editors
therapeutic to improve attention function in children with ADHD. FDA. Available
P. A. Hancock and N. Meshkati (North-Holland), 139–183. doi:10.1016/S0166-
online at: https://www.fda.gov/news-events/press-announcements/fda-permits-
4115(08)62386-9
marketing-first-game-based-digital-therapeutic-improve-attention-function-children-
adhd (Accessed November 19, 2024). Hatem, S. M., Saussez, G., della Faille, M., Prist, V., Zhang, X., Dispa, D., et al. (2016).
Rehabilitation of motor function after stroke: a multiple systematic review focused on
Cramer, S. C., Nelles, G., Benson, R. R., Kaplan, J. D., Parker, R. A., Kwong, K. K., et al.
techniques to stimulate upper extremity recovery. Front. Hum. Neurosci. 10, 442. doi:10.
(1997). A functional MRI study of subjects recovered from hemiparetic stroke. Stroke
3389/fnhum.2016.00442
28, 2518–2527. doi:10.1161/01.STR.28.12.2518
Henrique, P. P. B., Colussi, E. L., and De Marchi, A. C. B. (2019). Effects of exergame
da Silva Jaques, E., Figueiredo, A. I., Schiavo, A., Loss, B. P., da Silveira, G. H., Sangalli,
on patients’ balance and upper limb motor function after stroke: a randomized
V. A., et al. (2023). Conventional mirror therapy versus immersive virtual reality mirror
controlled trial. J. Stroke Cerebrovasc. Dis. 28, 2351–2357. doi:10.1016/j.
therapy: the perceived usability after stroke. Stroke Res. Treat. 2023, 1–8. doi:10.1155/
jstrokecerebrovasdis.2019.05.031
2023/5080699
Hoermann, S., Ferreira dos Santos, L., Morkisch, N., Jettkowski, K., Sillis, M., Devan,
Darbois, N., Guillaud, A., and Pinsault, N. (2018). Do robotics and virtual reality add
H., et al. (2017). Computerised mirror therapy with augmented reflection technology for
real progress to mirror therapy rehabilitation? A scoping review. Rehabilitation Res.
early stroke rehabilitation: clinical feasibility and integration as an adjunct therapy.
Pract. 2018, 1–15. doi:10.1155/2018/6412318
Disabil. Rehabilitation 39, 1503–1514. doi:10.1080/09638288.2017.1291765
De Luca, R., Gangemi, A., Maggio, M. G., Bonanno, M., Calderone, A., Mazzurco
Horne, M., Thomas, N., McCabe, C., Selles, R., Vail, A., Tyson, S., et al. (2015).
Masi, V. M., et al. (2024). Effects of virtual rehabilitation training on post-stroke
Patient-directed therapy during in-patient stroke rehabilitation: stroke survivors’ views
executive and praxis skills and depression symptoms: a quasi-randomised clinical trial.
of feasibility and acceptability. Disabil. Rehabilitation 37, 2344–2349. doi:10.3109/
Diagnostics 14, 1892. doi:10.3390/diagnostics14171892
09638288.2015.1024341
De Pasquale, P., Bonanno, M., Mojdehdehbaher, S., Quartarone, A., and Calabrò, R. S.
Hsu, H.-Y., Kuo, L.-C., Lin, Y.-C., Su, F.-C., Yang, T.-H., and Lin, C.-W. (2022).
(2024). The use of head-mounted display systems for upper limb kinematic analysis in
Effects of a virtual reality–based mirror therapy program on improving sensorimotor
post-stroke patients: a perspective review on benefits, challenges and other solutions.
function of hands in chronic stroke patients: a randomized controlled trial.
Bioengineering 11, 538. doi:10.3390/bioengineering11060538
Neurorehabil Neural Repair 36, 335–345. doi:10.1177/15459683221081430
Deakin, A., Hill, H., and Pomeroy, V. M. (2003). Rough guide to the fugl-meyer
Iosa, M., Galeoto, G., De Bartolo, D., Russo, V., Ruotolo, I., Spitoni, G. F., et al. (2021).
assessment: upper limb section. Physiotherapy 89, 751–763. doi:10.1016/S0031-
Italian version of the Pittsburgh rehabilitation participation scale: psychometric analysis
9406(05)60502-0
of validity and reliability. Brain Sci. 11, 626. doi:10.3390/brainsci11050626

Frontiers in Bioengineering and Biotechnology 19 frontiersin.org

De Pasquale et al. 10.3389/fbioe.2025.1655416

Ivanova, E., Minge, M., Schmidt, H., Thüring, M., and Krüger, J. (2017). User- telemedicine approach. J. Rehabilitation Med. 41, 1016–1102. doi:10.2340/16501977-
centered design of a patient’s work station for haptic robot-based telerehabilitation after 0459
stroke. Curr. Dir. Biomed. Eng. 3, 39–43. doi:10.1515/cdbme-2017-0009
Pollock, A., Farmer, S. E., Brady, M. C., Langhorne, P., Mead, G. E., Mehrholz, J., et al.
Kim, J.-H. (2017). The effects of training using EMG biofeedback on stroke patients (2014). Interventions for improving upper limb function after stroke. Cochrane
upper extremity functions. J. Phys. Ther. Sci. 29, 1085–1088. doi:10.1589/jpts.29.1085 Database Syst. Rev. 2014, CD010820. doi:10.1002/14651858.CD010820.pub2
Kwakkel, G., van Peppen, R., Wagenaar, R. C., Wood Dauphinee, S., Richards, C., Ramachandran, V. S., and Rogers-Ramachandran, D. (1996). Synaesthesia in
Ashburn, A., et al. (2004). Effects of augmented exercise therapy time after stroke. Stroke phantom limbs induced with mirrors. Proc. Biol. Sci. 263, 377–386. doi:10.1098/
35, 2529–2539. doi:10.1161/01.STR.0000143153.76460.7d rspb.1996.0058
Laffranchi, M., D’Angella, S., Vassallo, C., Piezzo, C., Canepa, M., De Giuseppe, S., Rand, D., Givon, N., and Avrech Bar, M. (2018). A video-game group intervention:
et al. (2021). User-centered design and development of the modular TWIN lower limb experiences and perceptions of adults with chronic stroke and their therapists:
exoskeleton. Front. Neurorobotics 15, 709731. doi:10.3389/fnbot.2021.709731 intervention de groupe à l’aide de jeux vidéo: expériences et perceptions d’adultes
en phase chronique d’un accident vasculaire cérébral et de leurs ergothérapeutes. Can.
Laver, K. E., Lange, B., George, S., Deutsch, J. E., Saposnik, G., and Crotty, M. (2017).
J. Occup. Ther. 85, 158–168. doi:10.1177/0008417417733274
Virtual reality for stroke rehabilitation. Cochrane Database Syst. Rev. 2018. doi:10.1002/
14651858.CD008349.pub4 Ríos-Hernández, M., Jacinto-Villegas, J. M., Portillo-Rodríguez, O., and Vilchis-
González, A. H. (2021). User-centered design and evaluation of an upper limb
Lee, S. W., Wilson, K. M., Lock, B. A., and Kamper, D. G. (2011). Subject-specific
rehabilitation system with a virtual environment. Appl. Sci. 11, 9500. doi:10.3390/
myoelectric pattern classification of functional hand movements for stroke survivors.
app11209500
IEEE Trans. Neural Syst. Rehabil. Eng. 19, 558–566. doi:10.1109/TNSRE.2010.2079334
Sadarangani, G. P., Jiang, X., Simpson, L. A., Eng, J. J., and Menon, C. (2017). Force
Li, S., Liu, J., Bhadane, M., Zhou, P., and Rymer, W. Z. (2014). Activation deficit
myography for monitoring grasping in individuals with stroke with mild to moderate
correlates with weakness in chronic stroke: evidence from evoked and voluntary EMG
upper-extremity impairments: a preliminary investigation in a controlled environment.
recordings. Clin. Neurophysiol. 125, 2413–2417. doi:10.1016/j.clinph.2014.03.019
Front. Bioeng. Biotechnol. 5, 42. doi:10.3389/fbioe.2017.00042
Liarokapis, M. V., Artemiadis, P. K., Kyriakopoulos, K. J., and Manolakos, E. S.
Saleh, S., Yarossi, M., Manuweera, T., Adamovich, S., and Tunik, E. (2017). Network
(2013). A learning scheme for reach to grasp movements: on EMG-based interfaces
interactions underlying mirror feedback in stroke: a dynamic causal modeling study.
using task specific motion decoding models. IEEE J. Biomed. Health Inf. 17, 915–921.
Neuroimage Clin. 13, 46–54. doi:10.1016/j.nicl.2016.11.012
doi:10.1109/JBHI.2013.2259594
Sarasola-Sanz, A., Irastorza-Landa, N., López-Larraz, E., Shiman, F., Spüler, M.,
Maggio, M. G., Torrisi, M., Buda, A., De Luca, R., Piazzitta, D., Cannavò, A., et al.
Birbaumer, N., et al. (2018). Design and effectiveness evaluation of mirror myoelectric
(2020). Effects of robotic neurorehabilitation through lokomat plus virtual reality on
interfaces: a novel method to restore movement in hemiplegic patients. Sci. Rep. 8,
cognitive function in patients with traumatic brain injury: a retrospective case-control
16688. doi:10.1038/s41598-018-34785-x
study. Int. J. Neurosci. 130, 117–123. doi:10.1080/00207454.2019.1664519
Sarasola-Sanz, A., López-Larraz, E., Irastorza-Landa, N., Rossi, G., Figueiredo, T.,
Maggio, M. G., Cezar, R. P., Milardi, D., Borzelli, D., De Marchis, C., D’Avella, A., et al.
McIntyre, J., et al. (2022). Real-time control of a multi-degree-of-freedom mirror
(2023). Do patients with neurological disorders benefit from immersive virtual reality? A
myoelectric interface during functional task training. Front. Neurosci. 16, 764936.
scoping review on the emerging use of the computer-assisted rehabilitation environment.
doi:10.3389/fnins.2022.764936
Eur. J. Phys. Rehabil. Med. 60, 37–43. doi:10.23736/S1973-9087.23.08025-5
Semprini, M., Lencioni, T., Hinterlang, W., Vassallo, C., Scarpetta, S., Maludrottu, S.,
Mani Bharathi, V., Manimegalai, P., George, S. T., Pamela, D., Mohammed, M. A.,
et al. (2022). User-centered design and development of TWIN-Acta: a novel control
Abdulkareem, K. H., et al. (2024). A systematic review of techniques and clinical
suite of the TWIN lower limb exoskeleton for the rehabilitation of persons post-stroke.
evidence to adopt virtual reality in post-stroke upper limb rehabilitation. Virtual Real.
Front. Neurosci. 16, 915707. doi:10.3389/fnins.2022.915707
28, 172. doi:10.1007/s10055-024-01065-1
Seo, G., Kishta, A., Mugler, E., Slutzky, M. W., and Roh, J. (2022). Myoelectric
Massetti, T., da Silva, T. D., Crocetta, T. B., Guarnieri, R., de Freitas, B. L., Bianchi Lopes,
interface training enables targeted reduction in abnormal muscle co-activation.
P., et al. (2018). The clinical utility of virtual reality in neurorehabilitation: a systematic
J. NeuroEngineering Rehabilitation 19, 67. doi:10.1186/s12984-022-01045-z
review. J. Cent. Nerv. Syst. Dis. 10, 1179573518813541. doi:10.1177/1179573518813541
Song, R., Tong, K., Hu, X., and Zhou, W. (2013). Myoelectrically controlled wrist
Mekbib, D. B., Debeli, D. K., Zhang, L., Fang, S., Shao, Y., Yang, W., et al. (2021). A
robot for stroke rehabilitation. J. NeuroEngineering Rehabilitation 10, 52. doi:10.1186/
novel fully immersive virtual reality environment for upper extremity rehabilitation in
1743-0003-10-52
patients with stroke. Ann. N. Y. Acad. Sci. 1493, 75–89. doi:10.1111/nyas.14554
Thieme, H., Mehrholz, J., Pohl, M., Behrens, J., and Dohle, C. (2012). Mirror therapy
Meyer, J. T., Schrade, S. O., Lambercy, O., and Gassert, R. (2019). “User-centered
for improving motor function after stroke. Cochrane Database Syst. Rev. 2012. doi:10.
design and evaluation of physical interfaces for an exoskeleton for paraplegic users,” in
1002/14651858.CD008449.pub2
2019 IEEE 16th International Conference on Rehabilitation Robotics (ICORR),
Toronto, ON, Canada (IEEE), 1159–1166. Torrisi, M., Maggio, M. G., De Cola, M. C., Zichittella, C., Carmela, C., Porcari, B.,
et al. (2021). Beyond motor recovery after stroke: the role of hand robotic rehabilitation
Morone, G., Spitoni, G. F., Bartolo, D. D., Ghooshchy, S. G., Iulio, F. D., Paolucci, S.,
plus virtual reality in improving cognitive function. J. Clin. Neurosci. 92, 11–16. doi:10.
et al. (2019). Rehabilitative devices for a top-down approach. Expert Rev. Med. Devices
1016/j.jocn.2021.07.053
16, 187–195. doi:10.1080/17434440.2019.1574567
Triandafilou, K. M., Tsoupikova, D., Barry, A. J., Thielbar, K. N., Stoykov, N., and
Morone, G., Papaioannou, F., Alberti, A., Ciancarelli, I., Bonanno, M., and Calabrò, R.
Kamper, D. G. (2018). Development of a 3D, networked multi-user virtual reality
S. (2024). Efficacy of sensor-based training using exergaming or virtual reality in
environment for home therapy after stroke. J. NeuroEngineering Rehabilitation 15, 88.
patients with chronic low back pain: a systematic review. Sensors 24, 6269. doi:10.3390/
doi:10.1186/s12984-018-0429-0
s24196269
Tsekleves, E., Paraskevopoulos, I. T., Warland, A., and Kilbride, C. (2016).
Mubin, O., Alnajjar, F., Al Mahmud, A., Jishtu, N., and Alsinglawi, B. (2022).
Development and preliminary evaluation of a novel low cost VR-based upper limb
Exploring serious games for stroke rehabilitation: a scoping review. Disabil. Rehabil.
stroke rehabilitation platform using wii technology. Disabil. Rehabilitation Assistive
Assist. Technol. 17, 159–165. doi:10.1080/17483107.2020.1768309
Technol. 11, 413–422. doi:10.3109/17483107.2014.981874
Nissler, C., Nowak, M., Connan, M., Büttner, S., Vogel, J., Kossyk, I., et al. (2019).
Veerbeek, J. M., van Wegen, E., van Peppen, R., van der Wees, P. J., Hendriks, E.,
VITA-an everyday virtual reality setup for prosthetics and upper-limb rehabilitation.
Rietberg, M., et al. (2014). What is the evidence for physical therapy poststroke? A
J. Neural Eng. 16, 026039. doi:10.1088/1741-2552/aaf35f
systematic review and meta-analysis. PLoS ONE 9, e87987. doi:10.1371/journal.pone.
Patel, G. K., Nowak, M., and Castellini, C. (2017). Exploiting knowledge composition 0087987
to improve real-life hand prosthetic control. IEEE Trans. Neural Syst. Rehabil. Eng. 25,
Zhang, X., and Zhou, P. (2012). High-density myoelectric pattern recognition toward
967–975. doi:10.1109/TNSRE.2017.2676467
improved stroke rehabilitation. IEEE Trans. Biomed. Eng. 59, 1649–1657. doi:10.1109/
Perez-Marcos, D., Bieler-Aeschlimann, M., and Serino, A. (2018). Virtual reality as a TBME.2012.2191551
vehicle to empower motor-cognitive neurorehabilitation. Front. Psychol. 9, 2120. doi:10.
Zirbel, C., Zhang, X., and Hughes, C. (2018). “The VRehab system: a low-cost Mobile
3389/fpsyg.2018.02120
virtual reality system for post-stroke upper limb rehabilitation for medically
Piron, L., Turolla, A., Agostini, M., Zucconi, C., Cortese, F., Zampolini, M., et al. underserved populations,” in 2018 IEEE Global Humanitarian Technology
(2009). Exercises for paretic upper limb after stroke: a combined virtual-reality and Conference (GHTC), San Jose, CA, USA, 18-21 October 2018 (IEEE), 1–8.

Frontiers in Bioengineering and Biotechnology 20 frontiersin.org