Neurofunctional correlates of a neurorehabilitation system based on eye movements in chronic stroke impairment levels: A pilot study

Abstract Introduction Rehabilitation after a stroke is widely considered fundamental to improve secondary functional impairments. Accessible methods based on motor learning, motor transfer and virtual environments are necessary to help to improve stroke patients’ quality of life. Objectives Continuing the line of our previous studies, this work investigated the effect of our new and innovative game‐based virtual reality training using the control of virtual objects with gaze in three chronic stroke survivors. Methods All participants performed an eye‐controlled virtual training task for 4 weeks. Pre‐ and post‐training evaluation were carried out with the Fugl‐Meyer Assessment for upper extremity scale as well as performing a tracking task inside an MRI scanner with a MRI‐compatible eye‐tracker or a joystick. Results Neural results for each participant show the increase of activity in the motor cortex, basal ganglia and cerebellum for both effectors (hand or eye). Conclusion These promising results have a potential application as a new game‐based neurorehabilitation approach to enhance the motor activity of stroke patients.

the affected limbs which is not always possible because of the patients' limitations (Camporesi et al., 2013). It is, therefore, necessary to create new accessible methods for all stroke survivors.
In the field of neurorehabilitation, it is important to find proper strategies to increase cerebral activity that could promote a better functional recovery (Crofts et al., 2020). Considering this principle, it is necessary to bear in mind that it would be necessary to use motor learning mechanisms to promote neuroplasticity or motor transfer, which are fundamental concepts in neurorehabilitation. Neuroplasticity consists of reorganizing and structuring brain function in response to injuries and stimuli (Su & Xu, 2020) and has been considered useful in the rehabilitation of stroke survivors and is highly dependent on motor learning (Dayan & Cohen, 2011;Dimyan & Cohen, 2011;Small et al., 2013). Furthermore, motor transfer, which is based on the application of a learned skill in a different context or task, is one of the major outcomes of motor learning (Censor, 2013). On the other hand, the use of virtual reality can provide training opportunities for many stroke patients to improve their adherence to treatment (Brunner et al., 2016;Krakauer, 2006), through more repetitive and engaging training that could also enhance neuroplasticity (Kim et al., 2020).
Based on the issues mentioned above, an approach to neurorehabilitation has been developed by the authors based on training eye control of virtual objects, which has been tested on healthy volunteers (García-Ramos et al., 2021;Gebert et al., 2017;Modroño et al., 2020;Modroño et al., 2015;Modroño et al., 2013). The objective of this study is to explore the effect of this new game-based ocular virtual reality training in three different profiles of stroke survivors.
Based on the results obtained in healthy volunteers, it can be hypothesized that: (a) virtual training with ocular movements in stroke participants should lead to an increase of cerebral activity in sensorimotor regions that (b) can be associated with an increase in the accuracy in eye movements, which can be transferred to the hand. This virtual training could be a useful tool in neurorehabilitation to increase cerebral activity by activating motor systems to help to improve recovery in stroke survivors.

Participants
Three chronic stroke survivors (1 female, 2 males) with different levels of impairment were recruited from the Association for Acquired

Motor and functional assessment
An initial evaluation was carried out to check the baseline of participants in the pretraining (Table 1). In order to do this, the researchers valued spasticity using the Modified Ashworth Scale (Bohannon & Smith, 1987), degree of dependence in daily activities assessed by the Modified Rankin Scale (van Swieten et al., 1988), the capacity to execute basic everyday activities with the Spanish version of the Barthel Index (Baztán et al., 1993), grade of impairment caused by the stroke by the adapted version in Spanish of the National Institutes of Health Stroke Scale (NIHSS) (Domínguez, 2006), and evaluated conscious level by the Glasgow Coma Scale (Teasdale & Jennett, 1974 The motor domain for the upper limb has a maximum score of 66 points (Fugl-Meyer et al., 1975). This scale is highly recommended for clinical use in stroke research (Sullivan et al., 2013).

Case 1:
The first participant was a 61-year-old male, who had suffered an ischemic stroke of the left anterior circulation 13 months prior to the experiment ( Figure 1). He had a total degree of independence and capacity to execute basic activities in daily life. There was no presence of spasticity in the affected limbs and he had a maximum level of consciousness. In addition, the NIHSS showed a low degree of deterioration caused by the stroke. After the initial evaluation based on the motor and functional assessment scales, P1 (participant 1) was classified as having a low level of impairment.

Case 2:
The second participant was a 63-year-old male, who had suffered a hemorrhagic stroke of the right hemisphere in the occipital-parietal area 10 months before the experiment (

F I G U R E 2
Experimental environment where the participant practices with an eye-tracker controlled by ocular movements in a PC with a virtual task.

Case 3:
The third participant was a 78-year-old female, who had suffered an ischemic stroke of the left hemisphere in the occipital-parietal area 49 months prior to the experiment ( Figure 1). She was unable to attend to her bodily needs and to walk without assistance, which correlate with a severe degree of dependency according to her score in the Barthel Index. She presented a slight increase in tone during passive movements, no motor response and difficulties to communicate due to the use of inappropriate words. Moreover, she showed a severe degree of deterioration as measured by NIHSS. After the initial evaluation based on the motor and functional assessment scales, P3 (participant 3) was classified with a high level of impairment in this research.

Intervention task
The intervention was based on

fMRI scanning and MR image acquisition
During the fMRI scanning, which was performed on two different

fMRI task
The fMRI task consisted of the continuous tracking of a target moving horizontally in a sine-cosine waveform (Figure 3) in which the participant must try to pursue the target as closely as possible by using a sensor. A red circle (cursor) appeared on a black screen to track the gray circle. The experimental task had two conditions: fixation (participant must focus their gaze on a gray cross: that is, a basal condition) and tracking (tracking blocks). Custom software using Visual C# and DirectX was developed to implement the task. The same task was used and described in a previous neuroimaging study (Modroño et al., 2020).

2.6
Image preprocessing and analysis fMRI data processing was performed using SPM12 software (www.fil.ion.ucl.ac.uk/spm/). In the preprocessing step the func-tional images were spatially realigned, unwarped, normalized, and smoothed. Visual inspection was used to validate normalized images.
The normalized images of 2 × 2 × 2 mm were smoothed by a full width at half maximum (

Behavioral results
The FMA scale for the upper extremity (FMA-UE) showed an improvement of two points (pre: 59, post: 61) in P2 after 4 weeks' intervention.
There were no changes in the total scores of the other two participants.
P1 obtained the maximum score (pre: 66, post: 66) and P3 a low score and 13.60 points in P3, and increased 7 points in P1 ( Figure 4). Thus, these results show more accuracy in P1 and P2 (but not in P3) concerning eye-tracking and a slightly better accuracy for P2 and P3 (but not for P1) in hand-tracking.

Neural results
The neural results of the two contrasts (eye-tracking-post > eye-trackingpre and hand-tracking-post > hand-tracking-pre) reveal an increase of activations in cortical and noncortical regions in the three participants for both runs using their gaze or the joystick, including motor areas (Tables 2 and 3). Three representative slices were selected for each participant in order to show some of the activated brain regions ( Figure 5).

DISCUSSION
The present study explores the effect of the authors' innovative neurorehabilitation approach based on the use of gaze to control virtual can be associated with an increase in the accuracy in eye movements, which can be transferred to the hand.
In the FMA-UE scale, one participant (P2) increased the total score by two points after the training period. This change took place in the subsection of coordination and speed. The other two participants had the same total score as in the pretraining evaluation, which was to be expected since P1 obtained the maximum score in the pretraining evaluation and P3 was not able to move her arm. It should be noted that in the context of the MCID (Minimum Clinically Important Difference), the increase of the score of P2 for the FMA-UE did not achieve a change equal or superior to four points to be considered clinically relevant (Page et al., 2012;Lundquist & Maribo, 2016), unlike other studies on similar neurorehabilitation systems, which were based, for example, on virtual reality and planar motion exercises , arm support and rehabilitation games (Prange et al., 2015), or robot-assisted rehabilitation (Ranzani et al., 2020). The differences in clinical relevance between such works and ours can be due to the sample size, and, for this reason, it would be necessary to perform new studies with more participants to clarify the clinical effect of our virtual training approach on the FMA-UE scale.
After the practice with the eye-controlled training games task, the accuracy was increased (that is, the MAE decreased, Figure 4a) for the eye-controlled fMRI task in two of the three participants (meanwhile a slight accuracy decrease was found for the other one). As regards the hand-controlled fMRI task, the accuracy was slightly increased ( Figure 4b) also for two of the three participants (meanwhile the accuracy was slightly decreased for another participant). The most noteworthy behavioral results are those obtained for participant 2, who had a middle level of impairment, and showed improvements in all the assessments (FMA-UE and MAE for eye and hand-controlled modalities). Taking together, these results seem to indicate that the practice with the eye-controlled training games could have led to a better performance in the eye-controlled sensorimotor task (that is, intertask transfer of motor learning) and, to a lesser extent, in the handcontrolled task (that is, intertask and intereffector transfer of motor learning). Here it should be noted that this is an exploratory study, and caution must be taken when making interpretations; however, the present results appear to be partially consistent with other studies TA B L E 2 Main anatomical structures associated with motor activity and activated by the effect of ocular virtual training in the run performed with the eye (Contrast: eye-tracking-post > eye-tracking-pre). Left superior frontal gyrus and supplementary motor area 6 13 -12 6 62 3.04

BA Cluster size (voxels) [X Y X] Peak T value
The MNI coordinates of peak activity are represented in each cluster. Threshold p < .05. FDR corrected al the voxel level.
in which the participants used the hand as an effector in a sensorimotor training task and obtained improvements (Ewolds et al., 2017;Lang et al., 2013). Furthermore, this study is a step forward to continuing the line of our previous research with healthy volunteers where improvements in hand-tracking accuracy were observed after a period of training in an eye-controlled task (Modroño et al., 2020).  : hand-tracking-post > hand-tracking-pre > hand-tracking-pre) in the eye-controlled and in the hand-controlled runs show neural activity increases after the training period, particularly in the premotor cortex, primary motor area, cerebellum, basal ganglia, putamen, and insula. This is also consistent with other fMRI research works where the use of virtual reality training or physical therapy was found to increase brain activity in similar regions (Ackerley et al., 2016;Hardwick et al., 2015;Mekbib et al., 2021;Vourvopoulos et al., 2019;Xiao et al., 2017).
The premotor cortex and the primary motor area are associated with motor control (Kantak et al., 2012;Ohbayashi, 2021;Vourvopoulos et al., 2019) and are strongly related to motor learning (Doyon & Benali, 2005). It should be mentioned here that some participants presented activations in the SMA (supplementary motor area), which has been related with different aspects of motor learning (Ruan et al., 2018), including intereffector motor transfer (Jung et al., 2019;Modroño et al., 2020;Perez et al., 2008;Perez et al., 2007) that could be affected in participants with motor disabilities such as stroke (Dirren et al., 2021). The most extended activations in this type of experiment appeared in the cerebellum, typically involved in the learning of sensorimotor tasks (Hardwick et al., 2013) and related with intermanual transfer tasks (Obayashi, 2004). Furthermore, the increase of activity in basal ganglia is associated with sensory motor training that could support casual changes in neuroplasticity in chronic stroke survivors (Zastron et al., 2019). The putamen and insula, in particular, play an important role in planning, executing, and learning motor skills (Doyon et al., 2009). In sum, an increase of activity in brain areas related to motor aspects with only the training on ocular movements in a virtual environment was observed in the present experiment.
The differences observed in the results might be due to the heterogeneity of the lesions of the participants. It should be noted that P1 was classified with a low level of impairment with a left ischemic stroke, P2 a medium level with a right hemorrhagic stroke and P3 a high level with a left ischemic stroke. The idea that motor recovery is a multifactorial process, which depends on multiple clinical factors such as lesion type, topography, size, and the baseline of somatosensory integrity after the stroke, is supported by other research works (Feydy et al., 2002;Ingemanson et al., 2019). Many factors could influence the recovery process for motor impairments; for example, the site of occlusion in the posterior circulation is associated with worse outcomes than in the anterior circulation (Sommer et al., 2018), the low baseline in NIHSS score is a predictor of good clinical results (Alexandre et al., 2021), or the kind of exercises used in rehabilitation, such as core exercises that are effective for improving motor recovery in stroke patients with severe motor impairments (Bacho & Khin, 2023).
The present study has some limitations. First, as a pilot study, the sample size is small which does not allow for the generalizability of the results. However, it is a valuable first approximation that allows for the assessment of the potential of this new neurorehabilitation tool. Second, the sample is heterogeneous concerning to the type of stroke and other clinical factors, which makes it difficult to compare the results of the participants. Nevertheless, it also may offer a unique perspective that can be built upon in future research. It should be noted that the study is a proof of concept, and further research with larger sample sizes and a control group is needed to confirm these first preliminary promising findings.

CONCLUSIONS
To the best of the authors' knowledge, this is the first time that stroke participants have performed an eye-controlled neurorehabilitation task. The present work shows that training on the ocular control of virtual objects can be a useful tool to increase neural activity in motor areas of interest that may be useful for neurorehabilitation in clinical practice and generate mechanisms to promote motor learning or motor transfer. This work expands our previous studies with healthy participants. The activations results obtained in the motor cortex, basal ganglia, and cerebellum of the three participants suggest that this new game-based neurorehabilitation approach has a promising potential application to help different types of stroke patients without resorting to limb movements. Therefore, it also invites us to continue this line of research by expanding the sample size.