Action Observation Training to Improve Upper Limb Function in Infants with Unilateral Brain Lesion – a Feasibility Study

ABSTRACT In this study, the feasibility of Action Observation Training (AOT) in combination with sensor-based measurements in infants at high risk of Unilateral Spastic Cerebral Palsy (UCP) were evaluated. Over a four-week period, eight infants at high risk of UCP performed AOT at home while wearing sensors with assistance of caregivers. Sensor data were compared to clinical assessments of upper limb function, the Hand Assessment for Infants (HAI) and the Mini-Assisting Hand Assessment (Mini-AHA). AOT training time and acceptance by the caregivers were considered as indicators for feasibility. The excellent training adherence and positive feedback of the caregivers showed that the AOT was feasible in this patient group and setting. Sensor measurements were accepted and displayed significant correlations with hand function. These preliminary results indicate the potential of wearable sensors to record upper limb function over the course of AOT for infants at high risk of UCP. Thus, AOT in combination with sensor measurements are proposed as a feasible training tool to complement usual care.


Introduction
Infants with unilateral brain lesions are at high risk of developing Unilateral Spastic Cerebral Palsy (UCP). [1][2][3] The causes of unilateral brain lesions include intraventricular hemorrhages, periventricular leukomalacia, arterial ischemic strokes, and venous infarctions. Infants with UCP often prefer to use only one hand, while the other hand is used less effectively. Consequently, infants with UCP have limitations of their bimanual hand function. Limitations of unilateral hand function in children with unilateral brain lesions may have a negative impact on quality of life. 4 The assessment of upper limb and hand function is essential for the detection of asymmetric hand use. This information is necessary for the planning and evaluation of interventions. Several methods for use in children with UCP have been evaluated in clinical settings. 5 The Hand Assessment for Infants (HAI) 6 and the Mini-Assisting Hand Assessment (Mini-AHA) 7 are well-established assessment tools based on observation of hand skills to measure how infants use both their hands separately and together during play. 6,7 In recent years, sensor technology has been used for objective, quantitative evaluations of movements in clinical contexts and sports. 8 Wearable sensors are compact, lightweight, and inexpensive and typically include accelerometers and gyroscopes. Previous studies have used them to assess arm movement asymmetries in children from three years upwards and in young adults with UCP. 9,10 The flexibility and compactness of wearable sensors allow its use also outside the clinical environment, for example in people's homes. Although home measurements tend to be less controlled than those obtained in a hospital setting, 11 they in return can be implemented in a natural, familiar environment over extended periods of time and across a wide range of activities. To the best of our knowledge, wearable sensors have not yet been used in the evaluations of infants with unilateral brain lesions, neither in clinical settings nor at home.
Various interventions are used to improve bimanual hand function in children with UCP. 12 One is Action Observation Training (AOT), 13 which involves purposeful observations of meaningful actions and their imitation. The goals are to restore damaged cerebral networks and to support the acquisition of motor functions through the activation of mirror neurons. 14 AOT has shown promising results in improving upper limb function in adults after stroke 15 and in children with hemiparesis aged five years and older. 16,17 The feasibility of AOT in infants has not yet been studied. The ability to learn new skills through observation starts to develop at approximately nine to twelve months of age. 18 Thus, AOT should be a suitable training method in this age range that is worth testing in clinical practice.
This study had two objectives: a) to investigate whether AOT is a feasible and well-accepted intervention for infants with unilateral brain lesions aged nine to twelve months and their parents, and b) to investigate the usability of home-based long-term measurements with accelerometers as a quantitative adjunct to assessments based on observation. Five hypotheses were based on these two objectives. We assumed that (i) parents and their infants completed at least 70% of the recommended AOT training time. (ii) Parents believe AOT is feasible and accept it as an intervention. (iii) Accelerometers are usable with infants. (iv) Measurements with accelerometers on both upper arms are feasible with infants. (v) Accelerometery measurements correlate with the clinical measurements of upper limb function in infants.

Study Sample and Recruitment
The inclusion criteria were a) term or preterm infants who had sustained unilateral brain lesions (e.g. intraventricular cerebral hemorrhage or neonatal arterial ischemic stroke) and show clinical signs of UCP, b) aged nine to twelve months (corrected age for preterm infants), c) fully informed parental consent, and c) ability of the parent(s) to understand the protocol.
The participating parties consisted of the parents or legal guardians and the infant diagnosed with unilateral brain lesions.
Infants were not eligible if they had: a) diagnosed visual impairment, b) other diagnoses affecting upper limb function (e.g. plexus palsy), or c) ongoing involvement in another intervention study aiming to improve hand function. Based on data from the Swiss Neuropediatric Stroke Registry and the University Children's Hospital Bern, around 20 newborns are diagnosed with unilateral brain lesions every year. Therefore, the targeted sample size was between five and ten participants.
Patients were recruited between March 2020 and May 2021 at the University Children's Hospital Bern. There, all preterm infants under 32 weeks gestational age and all infants with cerebral lesions who were cared for in the neonatology department were followed up. Additionally, infants directly referred to the occupational therapy and physiotherapy departments were screened.
Before enrolling in the study, the parents of each infant gave written informed consent. The trial was approved by the Bern Cantonal Ethics Committee (Project ID 2019-02028) and conducted in accordance with the latest version of the Declaration of Helsinki. The study was registered on ClinicalTrials.gov (NCT04194281).

Study Design
The trial was conducted as a pre -post intervention study. An uncontrolled design was appropriate, as we aimed to assess the feasibility of implementing AOT in combination with sensorbased measurements in infants at high risk of UCP. 19

Procedure and Data Collection
At the first appointment in the clinic, the children were fitted with the accelerometers and their caregivers' received instructions on their usage. The infants should wear the sensors twice a week during waking hours for a total of twelve days, which corresponds to a total study duration of six weeks. During a "sensor-day," the caregivers had to put on the sensors when the infant woke up and to remove them before they went to nocturnal sleep. The sensors were not allowed to be removed for day naps, except for baths and other waterrelated activities.
After the fitting of the accelerometers, the Hand Assessment for Infants (HAI) 6 and Mini-Assisting Hand Assessment (Mini-AHA) 7 were performed (pretest). In addition, the infants' ability to handle objects during everyday activities was described by the Mini-Manual Ability Classification System (Mini-MACS). 20 After analysis of the pretest-assessments HAI and Mini-AHA, the second appointment took place after four "sensordays" (two weeks) either in the clinic or at the participants' home, where the caregivers received feedback on the pretest results. Additionally, instructions for the now starting AOT (four weeks) and for usage of the therapy diary were given.
At the end of the study after six weeks, the HAI and Mini-AHA were repeated (posttest) in the clinic while the children were wearing the sensors. Additionally, a questionnaire on the feasibility and acceptance of the AOT, the System Usability Scale (SUS), and a sensor-specific questionnaire were given to the parents to obtain their feedback. The questionnaires were filled out at the parents' home and returned by mail. To determine adherence to the accelerometers, the total daily wearing time of the sensors was evaluated.

Intervention
The goal of Action Observation Training [AOT] is to support the learning of motor functions of the affected upper limb and the bimanual hand function by activating mirror neurons. 21 Compared to other trainings to improve hand function (e.g. Constraint-induced movement therapy) the core content of AOT is to observe arm, hand and finger movements of others. Because each infant's skill level was based on the evaluation of the HAI and Mini-AHA (pretest), individual goals for the AOT were set for this study (e.g. "grasping," "moves fingers," "moves upper arm," "bimanual manipulation") and discussed with a caregiver. To achieve these individual goals, the therapist prepared and lent the caregiver(s) a selection of appropriate toys, materials and movement sequences that match the child's current level of motor function (for an example see Figure 1). The caregivers were instructed to complete the materials with their own toys and materials based on the infants' interests.
Further, the therapist conducted a brief AOT demonstration session with the infant and prepared written instructions for the caregiver. These included suggestions for training implementation and for promoting the infant's motivation, attention, and interest. Caregivers were instructed to repeatedly demonstrate arm, hand, and finger movements and actions to their infant in a playful and age-appropriate manner. The movements can be performed with and without objects and one-handed or two-handed. During the observation, the child should sit in a stable position, as upright as possible, so that it can concentrate fully on the observation (e.g. in a high chair, on the lap of a person, in a stroller). The person demonstrating the movements can sit either opposite, behind or directly next to the child. The child should have a good view of the movements being demonstrated. The environment should be as distraction-free as possible so that the child can concentrate well on the movements being demonstrated.
After the demonstration, the infant had the opportunity to imitate the action or to play with the same toy for a moment before the demonstration continued. The caregiver and infants performed the AOT for four weeks, 20 minutes a day. They were advised to divide the daily training goal of approximately 20 minutes into several short sessions. The daily training time and duration of four weeks (total 560 minutes) are similar to other interventions at this age. 22 Outcome Measures Therapy Diary. To measure adherence, a therapy diary was completed by the parents (supplemental material 1). There, the daily duration of training and the number of sessions per day were recorded. The caregiver who had performed the AOT session with the infant additionally noted the mood of the infant during the training.

Questionnaire on the Feasibility and Acceptance of the AOT
A questionnaire was developed to examine the feasibility (part I) and acceptance (part II) of the AOT (supplemental material 2). It was based on similar questionnaires used in feasibility studies of interventions to improve upper extremity function in children with UCP. [23][24][25] The questionnaire was completed by the parent who conducted the most training sessions. A mean total score of≥70% was interpreted as positive feasibility (part I) and acceptance (part II). The interpretation of positive feasibility and acceptance is based on the interpretation of the SUS, where≥70% is also interpreted as positive. 26

Hand Assessment for Infants (HAI)
The HAI is an assessment of the quality of goal-directed unimanual and bimanual actions in infants aged three to twelve months. 6 A play situation is videotaped so that the items can be scored by a certified rater. In this study, the rater was an experienced occupational therapist of the Children's Hospital Bern and unblinded to the study purpose. The raw scores of the unimanual items range from 0-24 points for each hand. The raw score for the use of both hands together ranges from 0-58 points. Based on these two scores HAI units (0-100) are calculated, where a higher score indicates better hand function. For our feasibility study the HAI "Affected Hand score," the "Asymmetry Index" and the "Both Hand Measure" were evaluated.
The HAI shows excellent reliability. 27 The smallest detectable change (SDD) for the "Each Hand Score" is two points, and for the "Both Hand Measure" three HAI units. 27 The HAI also predicts UCP with high accuracy. 28

Mini-Assisting Hand Assessment (Mini-AHA)
The Mini-AHA validly measures and describes how effectively young children with UCP use their affected hand in bimanual play situations. 7 This assessment is used for infants aged eight to 18 months old. The therapist places the infant securely in an infant chair at a table and offers them a variety of ageappropriate toys. The toys have been specially selected to match objects the infant usually bimanually plays with. The play situation is videotaped and the same rater who scored the HAI scores the items. The person is certified for the Mini-AHA. Raw scores (20-80 points) are converted to Mini-AHA units (0-100), with 100 units representing optimal use of the affected hand. The Mini-AHA does not measure age-related developmental change, but rather differences in ability due to impairment. To date, there is no evidence on reliability and responsiveness to change.

Mini-Manual Ability Classification System (Mini-MACS)
The Mini-MACS classifies children's manual abilities according to five levels, ranging from level I (highest skills) to level V (most limited skills). Although the Mini-MACS was developed for children aged from twelve months to four years, it was used in this study to describe the infants' manual abilities due to a lack of alternatives.

Accelerometry
For the duration of the intervention, the infants wore two triaxial accelerometer sensors (Axivity A×3, Axivity Ltd, Newcastle upon Tyne, United Kingdom) during the day. The sensors were also worn during the play sessions assessing the HAI and Mini-AHA (pre-and posttest). The accelerometers were both attached to the distal upper arm close to the elbow joint (left and right) with a custom-made band with a hook-and-loop closure (see Figure 2). To determine adherence, wearing times per day were evaluated. The goal was to have the sensors worn two days per week for six weeks. For data analysis, the three-dimensional signals from the accelerometers were filtered to eliminate noise, slow drift, and gravity effects. The quantitative movement velocity was calculated based on the filtered signal. The velocity was summarized over time windows of five minutes each (for more details of data analysis see Appendix). For each of these time windows, an asymmetry index (AI) was calculated as AI where V H is the mean velocity of the healthy arm and V I is the mean velocity of the affected arm for every 5-minute window. This asymmetry index ranges between À 1; þ1 ½ � and is based on the Edinburgh Inventory formula. 29 An AI of 1 means that only the healthy arm is moved. An AI of − 1 means that only the affected arm is moved. An AI of 0 means symmetrical, i.e. healthy movements.

System Usability Scale (SUS)
To evaluate the usability of the accelerometers, nine out of the ten questions of the SUS were used. 30 As the official German translation of the sixth question ("I thought there was too much inconsistency in this system.") was difficult to understand and not relevant to answering the research question, it was omitted. The questionnaire was completed by the parents. A SUS score of≥70% was interpreted as good usability. 26

Sensor-Specific Questionnaire
To evaluate the practicability of measurements with accelerometers, a questionnaire adapted from another evaluation of accelerometers 31 was developed by the authors (supplemental material 3). The questionnaire was completed by the parents. A mean score of≥70% was interpreted as positive practicability. The interpretation is based on the interpretation of the SUS, where≥70% is interpreted as positive. 26

Data Analysis
The characteristics of the study group and the data from the diaries, questionnaires, and accelerometery were analyzed descriptively. Because of the small sample size, the nonparametric Wilcoxon test for paired data was used to test the difference between the pre-and posttest data from the HAI and Mini-AHA. To analyze the correlation of the accelerometery data (asymmetry index AI) obtained during pre-and posttests with the corresponding results of the HAI and Mini-AHA, Spearman's correlation coefficients r S were calculated and the effect sizes according to Cohen 27 reported. Pre-and posttests were treated as two separate measurements. For all statistical tests, a significance level of α ¼ 0:05 was chosen. Microsoft Office Excel 2019 and IBM® SPSS® Statistics version 25 were used for the analysis.

Recruitment and Characteristics of the Study Participants
Between March 2020 and May 2021, a total of 556 infants at increased risk for developmental disorders who had a followup-checkup were screened for the diagnosis of a unilateral brain lesion and clinical signs for UCP (Figure 3). Of the 39 infants with unilateral brain lesion, 27 had no clinical signs of UCP. The remaining twelve infants with clinical signs for UCP  and their parents were asked to participate in the study. Of these twelve families, four declined to participate. In two cases, the infants had other treatment priorities. For one family the distance to the hospital was too far and one family did not have sufficient time. Therefore, no reasons were because of the nature of the study intervention, not limiting adherence. There was no study dropout. Table 1 shows the clinical characteristics of the infants.
Due to the COVID-19 pandemic, the interventions could not begin until June 2020.

Therapy Adherence, Feasibility, and Acceptance of AOT
On average, a total training duration of 607 minutes (10.1 hours) was achieved (SD = 215 minutes (3.6 hours); median = 533 (8.9 hours); min -max = 415-1059 (6.9-17.7 hours)). A mean duration of training per day of 23 minutes was reported (SD = 11; median = 20; min -max = 0-60). On average, 1.4 training sessions were performed per day (SD = 1.0, median = 1, min-max = 0-5). The AOT feasibility questionnaire had a mean score of 23.1 out of 28 possible points (82.5%; SD = 3.76; median = 25; min -max = 18-27). The mean score for the acceptance questionnaire was 17.5 out of 24 points (72.9%; SD = 4.24; median = 16.5; min -max = 13-23). Seven out of eight parents agreed to continue implementing the therapy or aspects thereof in the future. All parents confirmed that the implementation of the AOT was easy and agreed with the statement that the intervention had a positive impact on bimanual hand use.

Hand Assessment for Infants and Mini-Assisting Hand Assessment
The HAI "Affected Hand Score," "Asymmetry Index" and "Both Hands Measure" scores and the Mini-AHA improved significantly from pre-to posttest ( Table 2). The effect size was r = −0.84 for all HAI scores and r = −0.90 for the Mini-AHA, corresponding to a strong effect. 32 The changes of the HAI "Affected Hand Score" and "Both Hands Measure" were both clinically meaningful (Figure 4). The infant with the highest pretest scores in the HAI and Mini-AHA (child-02) had very mild signs of UCP and their results are close to a Ceiling effect.

Accelerometery -Home-Based Measurements
One participating party did not properly adhere to the instructions given in the initial briefing; their data is reported separately as an outlier (child-08). The remaining seven participants wore the sensors for on average 13 days (SD = 3.1, median = 14; min -max = 8-18), 10.17 hours a day (SD = 1.9 h, median = 11.40 h; min -max = 3.00-13.25 h). The daily sensor wearing times are shown in Figure 5. The outlier participant wore the sensors for a total of 37 days for 2.93 hours a day (SD = 1.8 h, min -max = 5 min-2.74 h). The homebased measurements are shown in Figure 6. For infants with Mini-MACS I, the mean was close to symmetry, crossing the line multiple times, with the first quantile line reliably below AI = 0 and the third quantile line reliably above AI = 0. For infants with Mini-MACS II, the first quantile line crossed AI = 0 multiple times, but the mean velocity was mostly above the symmetry line at AI = 0. For infants with Mini-MACS III, the first quantile line was reliably above AI = 0.

Accelerometery -In-Clinic Measurements
One infant (child-04) did not wear the sensors during the posttest, only during the pretest. All other infants wore the sensors during the pre-and posttests, yielding n = 15 different measurements for the evaluation; eight pretests and seven posttests. There were significant correlations between the accelerometery-based asymmetry index AI and the HAI  For every participant, median wearing time is reported above the box. Child-08 did not adhere to the instructions and wore the sensors for shorter time periods, but for more days (37 days) than the other participants (mean = 12.5 days). Mean wearing times is reported both with and without child-08.

Discussion
Our study has shown that AOT is feasible for infants at high risk of UCP. Clinical assessment found that the infants' bimanual hand function improved. Furthermore, in-clinic measurements obtained through wearable sensors correlate strongly and significantly with the HAI and Mini-AHA assessments, and home-based measurements are clearly linked to the UCP symptom severity scale, the Mini-MACS.
Based on the results of the questionnaires, the hypothesis that AOT is feasible in infants and accepted by the parents was confirmed. Furthermore, therapy adherence was excellent, as was shown by the training times recorded in the diaries, confirming the second hypothesis, too. Because AOT can be performed in the home setting, the intervention has the potential to be made easily accessible to a broad population. However, it should also be noted that four of the twelve families approached declined to participate in the study due to different reasons. Whether the refusal of participation is related to the acceptance is unclear.
The results of the HAI and Mini-AHA showed a significant improvement of hand function with a large effect size. This is the first time that the hand function before and after an AOT has been investigated in infants at this age. A comparison of the effect size with studies of AOT in older age groups shows that these results are in line with or better than past findings. 21,22 Additionally, the change from HAI results from pre-to posttest correspond to a true change and is clinically meaningful. 27 The Mini-AHA results in pre-and posttest can also be interpreted as a true change. 7 The clinical important improvements suggest that further investigations are warranted, for example to explore whether these improvements sustained over time. However, the results must be interpreted with caution. It should be noted that the improvement from pre-to posttest may be due to reasons other than AOT, for example natural development 33 or other factors. It is also not known whether families implemented other treatments related to hand function at home. To control for such confounding factors, a randomized controlled design would be suitable for a future study.
From the therapists' point of view, the AOT was easy to perform with the infants. The participating infants were very curious and observed the demonstrated actions with great interest. By changing the demonstrated action and by showing a wide range of interesting objects and toys, attention could be sustained over a longer period of time. From the therapists' point of view the therapy instruction for the parents turned out to be relatively simple.
All but one of the accelerometery measurements in the clinic (pre-and posttest) could be made as planned. The accelerometery data correlated significantly with the scores of the HAI and Mini-AHA, confirming our hypothesis that clinical assessments would correlate with sensor-based measurements. Similar results have previously been reported in a study with older children and adults (three to 25 years old), where the MACS levels correlated with sensor-measured asymmetry. 10 Based on the answers to the questionnaires, we also confirmed the hypothesis that accelerometer-based measurements are feasible. The participants adhered to the planned wearing times at home, with individual deviations. The wearing times corresponded approximately to the waking hours that can be expected for nine to twelve month old infants. It is very likely that most of the relevant movements were recorded. Furthermore, the results of the questionnaires showed that the sensors were not experienced negatively. Willingness to use such a system over a longer period, i.e. up to six months, was confirmed by the parents of six of the eight infants. Knowing that the willingness to wear the sensors over a longer period of time is important so that measurements can be made over longer periods in future studies. This is very much in line with other long-term sensor studies. 34 However, the home-based measurements did not reflect the improvements that were shown by the Mini-AHA, HAI, and the accelerometer-based measurements in the pre-and posttest. Rather, the data appeared to reach a steady state during the six weeks of measurement. Since the home measurements were unsupervised, this less pronounced effect is to be expected. 34 Although home-based measurements allow insights into everyday life, measurement results tend to be less pronounced than the same measurements in a controlled environment, such as a clinical test. For the HAI and Mini-AHA, toys that require bimanual dexterity to play with are deliberately chosen. 7 In the home environment there could possibly be more barriers to using both hands as much as possible. Therefore, we confirm the hypothesis that accelerometerbased measurements at home are usable, but with reservations.
The parents of child-08 did not adhere to the home measurement protocol, resulting in shorter measurement sessions. Most likely this was caused by unclear communications or misunderstandings at the beginning of the study. In an ongoing therapy, this would have been noticed and corrected much earlier right after the first data segment evaluation. Nevertheless, for future studies a checkup phone call or a previous analysis could be planned. While the accelerometerbased measurements obtained from child-08 exhibited similar movement behavior to child-05, both with Mini-MACS II, their accelerometery results should be treated with caution as they show some inconstancy. Also, the HAI and Mini-AHA results of child-02 should be interpreted cautiously, as it showed almost a ceiling effect with the good scores due to the very mild UCP.
From the therapists' point of view, the application of the sensors during the clinical tests can be well implemented. The infants were not distracted at all. To date, the evaluation of the data requires appropriate software and the necessary expertise, so that it seems rather unrealistic in everyday clinical practice and is probably more applicable for research purposes.

Limitations of the Study and Future Research
As this study was an uncontrolled feasibility study, we were limited in correcting for confounding factors influencing clinical outcomes, such as usual care going on during the study intervention or motor improvements due to developmental changes. Future studies could correct for that by implementing a controlled design with a more homogenous sample.
Furthermore, the rater of the HAI and Mini-AHA was not blinded to the research objective, because she was involved in study planning and these outcomes were not the main focus.
It can be argued whether performing both assessments, the HAI and Mini-AHA, is necessary. The HAI presents an asymmetry index, which can be compared with accelerometer, and assesses the unaffected hand, while the Mini-AHA focuses on bimanual play only. To the best of our knowledge, there are no studies that have compared the two assessments. However, since it is a small age range where both assessments can be used (eight to twelve months), there is no reason not to use both complementary assessments to cover a wide range of hand function. For future studies, the HAI "Non-Affected Hand Score" should also be considered, because it could impact both upper limbs. Furthermore, it is recommended to start with the intervention directly after the pretest and not to wait two weeks as it was the case in this study. Further research should also include medium and long-term effects.
Another limitation is the use of a diary as an outcome measure of training duration. It is unclear how much time the infants spent observing actions and how much time they spent actively engaged in motor activity. If the movements and actions demonstrated were standardized as in other studies, the use of videos as an AOT demonstration tool would be a possibility. However, there is evidence that more activity can be measured in sensorimotor brain areas when a person is directly observed performing an action. 35 A possible means of standardization could be to give parents a uniform selection of toys and precise instructions for demonstrating the actions. Another limitation is the use of self-created questionnaires, which have not been standardized and tested before. Some questions were quite similar, for example: "The implementation of AOT was easy for me" and "It was easy to implement the AOT into everyday life" or the questions "Correctly donning the sensors was complicated." and "Attaching the motion sensors with the straps is awkward." In order to make the questionnaire more homogeneous and to simplify the data analysis, all questions could be formulated in the same way, positively or negatively. For future studies, the questionnaires should be revised.

Conclusion
AOT for infants nine to twelve months of age is a feasible intervention that can be implemented in daily life and is accepted by parents as a way to enhance bimanual hand function. The results indicate that the intervention has the potential to improve hand function in nine to twelve-month-olds with unilateral brain lesions. Further research is needed to investigate its efficacy.
As accelerometery measurements as a supplemental measurement tool are feasible and provide inferences about clinical values, they should be used as an adjunct to clinical outcome measures in future studies. They could provide more information about the movements of upper extremities across a wide range of activities in a familiar environment.