Abstract

Introduction:

Unilateral cerebral palsy (UCP) results from early brain injury, leading to motor impairments primarily affecting one upper limb. Action Observation Therapy (AOT), which engages the mirror neuron system to enhance motor function, represents a promising rehabilitative approach. However, its efficacy in UCP has been inconsistent across studies, possibly due to methodological differences or individual variability in brain damage. This observational longitudinal study within an intervention group aimed to explore whether structural brain lesions are associated with changes in motor dexterity and bimanual performance following AOT in children with UCP.

Methods:

Sixteen children with UCP underwent a structured AOT home training program. Motor function was assessed through the Box and Block Test (BBT) and the Assisting Hand Assessment (AHA) at four or five time points (T0, T1/T1_plus, T2, T3), totaling 64 evaluation sessions. Brain lesions were evaluated using the semi-quantitative magnetic resonance imaging (sqMRI) scale. Data were analyzed using linear mixed-effects models.

Results:

A significant increase in BBT scores for the dominant hand (BBT_dom) was observed at all post-treatment time points, while significant improvements in the non-dominant hand (BBT_non_dom) were observed at T2. AHA scores significantly increased at T1 and T2. Lesions involving the temporal lobe, thalamus, brainstem, and corpus callosum showed the strongest associations with reduced improvements in AHA and BBT_non_dom scores.

Discussion:

These findings suggest that structural brain lesions may influence the response to AOT in children with UCP. Further studies are required to confirm these results and to evaluate whether tailoring interventions according to neuroanatomical profiles could enhance rehabilitation outcomes in children with UCP.

1 Introduction

Cerebral palsy (CP) refers to a group of non-progressive disorders that impact movement and posture due to disturbances in early brain development (Bax et al., 2005). Unilateral spastic cerebral palsy (UCP), a common subtype, arises from congenital or early-acquired brain injury mainly affecting one hemisphere, leading to spasticity predominantly in the contralateral upper limb. Rehabilitation of motor functions is central to the therapeutic approach for UCP. It may involve interventions through intramuscular botulinum toxin A, upper-limb training, physical therapy, special devices, occupational therapy, and action observation training (AOT) (Buccino et al., 2012; Novak et al., 2020; Sakzewski et al., 2009). AOT uses systematic observation of meaningful actions followed by their execution to accelerate functional recovery in patients with motor impairments (Buccino et al., 2006). The theoretical foundation of AOT lies in the activation of the same neural structures during action observation as during the execution of those actions, a phenomenon derived from the “mirror neuron system” (Rizzolatti and Craighero, 2004). Imaging studies have shown that this system is crucial for action understanding and intention coding, imitation, and motor learning (Hari and Kujala, 2009), making it a promising target for therapeutic interventions in UCP.

Despite the promising theoretical underpinnings of AOT, empirical findings regarding its efficacy in improving motor functions in children with CP have been inconsistent (Buchignani et al., 2019). Several studies have reported significant benefits of AOT in enhancing motor skills in CP (Buccino et al., 2012; Jeong and Lee, 2020; Kim et al., 2014; Nuara et al., 2019; Quadrelli et al., 2019; Sgandurra et al., 2013; Simon-Martinez et al., 2025; Simon-Martinez et al., 2018). Indeed, a systematic review by Alamer and colleagues concluded that AOT is a promising intervention for upper limb rehabilitation in children with hemiplegic CP (Alamer et al., 2020). However, other studies have failed to confirm these findings, with some authors reporting no significant improvements (Kim, 2020; Kirkpatrick et al., 2016). Moreover, a meta-analysis by Abdelhaleem and colleagues (Abdelhaleem et al., 2021) found insufficient evidence to draw firm conclusions about the effectiveness of AOT in CP rehabilitation. These discrepancies could stem from methodological differences between studies or from clinical heterogeneity among treated patients, highlighting the need for identifying markers that could predict a patient’s response to AOT.

Among the markers that could predict AOT response, we can include neuroanatomical features. Several studies have demonstrated the potential of AOT to induce functional reorganization within motor-related brain areas, leading to improved motor outcomes in children with CP. The neural mechanisms underlying motor function, particularly the action observation network (AON) and the sensorimotor network (SMN) have been extensively studied in UCP. In a previous study conducted by our group, exploring the reorganization of AON and SMN in children with UCP using functional magnetic resonance imaging (fMRI), findings revealed that children with UCP showed higher lateralization of AON compared to typically developing (TD) peers, which was negatively correlated with clinical performance (Sgandurra et al., 2018). Consistently, in a pilot study performed by our group investigating the impact of AOT on brain reorganization in children with congenital hemiplegia using fMRI, we found that after three weeks of AOT, the experimental group showed a shift toward a more bilateral representation of the AON, suggesting that bilateral AON activation could be related with better motor performance in UCP. Moreover, the experimental group demonstrated significantly higher activation in the SMN compared to the control group, particularly in regions such as the ipsilesional primary motor area, pre-central gyrus, and supplementary motor area (Sgandurra et al., 2020). In the same line, the group of Buccino and colleagues (Buccino et al., 2018) found that children who underwent AOT exhibited significant improvements in motor function accompanied by stronger activation in the parietal-premotor circuit, particularly in the left premotor cortex, inferior frontal gyrus, and superior temporal gyrus.

Despite these insights, there is a notable gap in the literature regarding structural MRI characteristics that could predict the efficacy of AOT in children with UCP. Recently, Beani et al. (2024) observed moderate to strong associations between upper limb functional impairment and structural brain characteristics, as measured through a semiquantitative MRI scale (sqMRI) (Fiori et al., 2014). Specifically, associations between sqMRI scores contralateral to the more affected side and upper limb functional impairment, as measured through the Manual Ability Classification System (MACS) and the Box and Block Test (BBT), were identified. More severe brain injuries significantly correlated with poorer function in the non-dominant hand (Beani et al., 2024). These findings suggest that structural MRI markers could serve as potential predictors of response to rehabilitation strategies like AOT. The present study aims to explore whether structural brain lesions are associated with changes in motor dexterity and bimanual performance following AOT in children with UCP. This study could enable the identification of MRI markers that may be associated with the effect of AOT on motor function.

2 Materials and methods2.1 Participants

Children with UCP enrolled in the Tele-UPCAT project (Beani et al., 2023; Sgandurra et al., 2021) were selected for this study. Inclusion criteria were: (1) children, adolescents, or young adults aged 5 to 20 years; (2) confirmed having a diagnosis of UCP; (3) cognitive level within normal limits (IQ > 70); (4) availability to commit to a home program of intensive therapy for 3 weeks. Further inclusion criteria were the availability of a complete brain MRI exam to classify brain lesions according to different scales. Exclusion criteria were (1) orthopedic surgery or intramuscular botulinum toxin A injection in the upper limb prior to the starting of the rehabilitation session.

The upper limb function was assessed using specific outcome measures according to the study protocol. Specifically, the Box and Block Test (BBT) (Mathiowetz et al., 1985) measures unimanual dexterity in the activity domain, assessing both dominant (less affected) and non-dominant (more affected) hand separately. The Assisting Hand Assessment (AHA) (Krumlinde-sundholm et al., 2003) measures upper limb function during bimanual activities by evaluating spontaneous use of the assisting hand during a semistructured age-appropriated session.

The trial was approved by the Tuscan Pediatric Ethics Committee (number 169/2016) and registered at: http://www.clinicaltrials.gov (NCT03094455) on 16 March 2017. All parents provided written informed consent to participate in the trial.

2.2 Instruments2.2.1 Clinical hand function assessment2.2.1.1 Gross manual dexterity assessment

The Box and Block Test (BBT) is a standardized test used to measure gross manual dexterity, widely used with adults and later adopted also in pediatric age (Mathiowetz et al., 1985). It is quick, easy to administer, and cost-effective. The test apparatus consists of a box divided into two compartments by a central partition and 150 wooden blocks, each measuring 25 mm. Participants are instructed to move the blocks from one compartment to the other as quickly as possible within a one-minute timeframe. The number of blocks successfully moved is recorded, with a higher count indicating better manual dexterity. In the UCP cohort, the test is first performed with the dominant hand, followed by the non-dominant hand. A difference of 7 blocks was considered an indicator of clinically important change (Liang et al., 2021).

2.2.1.2 Assisting hand assessment

The Assisting Hand Assessment (AHA) is designed to evaluate how effectively children with unilateral upper limb dysfunction, such as obstetric brachial plexus palsy or UCP, use their non dominant hand during a semi-structured play session (Krumlinde-Sundholm et al., 2007). It is validated for children from 18 months up to 18 years (Louwers et al., 2016), with age-appropriated comparable sessions (from free play to different board games) in which objects requiring bimanual use are presented. After videotaping the session, the score is made by a certified scorer; the latest version, the 5.0, is composed of 20 items each scored with 4 points scale, which are processed based on Rasch analysis, returning a total sum score (1–80) converted also in AHA units score (1–100). Higher scores mean more symmetric hand use and better non-dominant side abilities. A difference of 5 AHA units is considered a clinically relevant change beyond measurement error.

2.2.2 Neuroimaging assessment

MRI data of all participants were acquired at the IRCCS Stella Maris Foundation, by using either a 1.5 T or 3 T MRI scanner (Signa Horizon 1.5 T, Signa Premier 3 T, GE Healthcare, Milwaukee, WI). The standard clinical protocol included two-dimensional T1-weighted, T2-weighted, and T2*-weighted- sequences, as well as three-dimensional T1-weighted, T2w FLAIR, and T2* Susceptibility weighted (SWI) sequences. The MR images were retrospectively evaluated using a reliable and validated semi-quantitative scoring system (Fiori et al., 2014; Fiori et al., 2015). An experienced pediatric neurologist (SF) assessed the images.

2.2.2.1 Brain lesion MRI assessment

Brain lesions on clinical MRI images were evaluated using a validated semi-quantitative (sqMRI) scoring system (Fiori et al., 2014). This method is based on a six-axial-slice template, with anatomical regions identified using specific MRI slices. The assessment systematically evaluates both hemispheric and subcortical structures on each hemisphere (right and left), generating subscores for the frontal, parietal, temporal, and occipital lobes; for subcortical structures including caudate, lenticular nuclei (putamen and globus pallidus), posterior limb of the internal capsule (PLIC), thalamus, and brainstem; as well as for the cerebellum and corpus callosum. Higher scores indicate greater involvement of the respective brain regions. Brain MRI images were also classified using the Surveillance of Cerebral Palsy in Europe (SCPE) classification system, based on the Magnetic Resonance Imaging Classification System (MRICS) (Himmelmann et al., 2017). The MRICS categorizes findings into five main groups: maldevelopments (Type I), predominant white matter injury (Type II), predominant gray matter injury (Type III), miscellaneous (Type IV), alongside a category for normal findings (Type V).

2.3 Procedure

Tele-UPCAT study was designed as an intention-to-treat clinical trial and implemented as a randomized AOT rehabilitation program and a Standard Care (SC) group, allocation-concealed, waitlist-controlled, and evaluator-blinded investigation. Participants enrolled after baseline evaluations, were randomized into either the immediate AOT or SC group. Participants underwent assessments at baseline (T0) and after a 3-week initial intervention period, which involved AOT for the experimental group and SC for the control group (T1). The SC group received typical care and after completing the SC phase, this group transitioned to the AOT intervention, followed by a secondary assessment upon completion (T1 plus). Both groups were subsequently evaluated at 8 weeks (T2) and 24 weeks (T3) post-intervention.

The intervention phase consisted of 15-day cycles of goal-oriented activities with increasing complexity over the training period (each daily session lasted approximately 1 h: alternating observation and action phases). The program began with eight unimanual training sessions, followed by seven bimanual sessions. The training was delivered through the Tele-UPCAT system (provided directly to the participant’s home), which included an all-in-one computer equipped with specialized software, a toy kit, and Actigraph devices.

During the Standard Care (SC) phase, participants continued their usual care without receiving any additional training. The Italian National Health System provides at least one or two therapy sessions per week, with a total duration of approximately 1 h a week, and this corresponded to the therapy received by all participants in our trial, as confirmed by the diaries completed by all families throughout the study. In contrast, the AOT intervention, as an intensive and structured home-based program delivered over 3 consecutive weeks (5 days per week, approximately 1 h per day), resulted in a substantially higher treatment intensity compared to SC.

Clinical assessments were conducted under controlled conditions by trained therapists, even when performed in the home setting. All outcome measures, including the Assisting Hand Assessment (AHA) and Box and Block Test (BBT), were video-recorded using standardized procedures. Subsequently, scoring was performed by certified assessors who were blinded to both group allocation and assessment time point. In particular, AHA evaluations were conducted by certified raters following standardized administration and scoring protocols, ensuring high reliability. The use of video recordings allowed for detailed and repeated evaluation, further enhancing scoring accuracy and consistency across time points.

2.4 Data analysis

To evaluate the effects of Action Observation Therapy (AOT) on motor dexterity and to explore structural MRI predictors of treatment response, we employed linear mixed-effects models (LMMs). We modeled the scores obtained from the BBT for both the dominant (BBT_dom) and non-dominant (BBT_non_dom) hands and the AHA scores as dependent variables. The models incorporated a random effect for each participant (denoted by “Code”) to capture individual variability. The fixed effects included time points (TIME: T0, T1/T1plus, T2, T3), Age, and Sex, enabling us to assess changes in motor dexterity throughout AOT while adjusting for potentially confounding factors. Indeed, several factors may impact recovery following early brain insult, including injury features (extent, severity, and location), age at the time of insult, and sex (Anderson et al., 2011). Moreover, the broad age range may introduce developmental variability; to account for this, age was included as a covariate in all mixed-effects models. Three primary models were developed: Model 1.1 with BBT_dom as the dependent variable, Model 1.2 with BBT_non_dom as the dependent variable, and Model 1.3 with AHA scores as the dependent variable.

To explore the contribution of structural MRI features to motor function over time, we developed secondary linear mixed-effects models by incorporating semi-quantitative MRI (sqMRI) scores as additional fixed effects. SqMRI variables were entered separately into each model and included scores for the frontal, temporal, parietal, and occipital lobes, as well as the caudate nucleus, lenticular nuclei, posterior limb of the internal capsule (PLIC), thalamus, and brainstem—classified as either ipsilesional or contralesional. Additional regions analyzed included the corpus callosum and cerebellum. All secondary models included TIME, age, sex, and lesion hemisphere (right vs. left) as fixed effects, and participant ID as a random effect. Moreover, as testing multiple brain regions increases the risk of type I error, p-values were adjusted using the false discovery rate (FDR) method, with statistical significance set at p < 0.05.

Given the high collinearity between sqMRI variables and lesion classification under the MRICS, the effect of MRICS was examined in separate mixed-effects models. Specifically, three models (one per outcome: BBT_dom, BBT_non_dom, and AHA) were fitted with MRICS classification (Type III vs. Type II) as a fixed effect, adjusting for TIME, age, and sex. MRICS Types I, IV, and V were excluded from analysis due to low representation in the sample. All analyses were conducted using R version 4.4.1.

3 Results

The study included 16 children with UCP assessed at time points (T0, T1, T1_plus T2, T3), totaling 64 evaluation sessions. Demographic and clinical characteristics are presented in Table 1 and Figure 1. Significant differences between patients with predominant Right vs. Left lesions were found in baseline AHA scores. The involvement of each brain region, as measured by sqMRI scores observed in the overall sample are summarized in the Supplementary Table S1.

VariableAll sample (n = 16)Right hemiplegia, left hemisphere lesion
(n = 9)Left hemiplegia, right_hemisphere lesion (n = 7)p-valueAge, (median [IQR])9.41 [8.55, 11.49]9.59 [9.04, 10.73]8.57 [7.04, 11.54]0.223Sex, male (%)8 (50)5 (55.6)3 (42.9)1Baseline AHA (median [IQR])48.50 [40.75, 58.50]41.00 [39.00, 55.00]53.00 [48.50, 66.00]0.039*Baseline BBT dominant (median [IQR])52.50 [46.75, 59.00]53.00 [48.00, 62.00]51.00 [46.50, 58.00]0.672Baseline BBT non-dominant (median [IQR])17.00 [12.75, 34.00]13.00 [11.00, 33.00]17.00 [16.00, 36.50]0.203GMFCS, N (%)0.294GMFCS_I13 (81.2)7 (100)GMFCS_II3 (18.8)3 (33.3)0GMFCS_III0 (0.0)0 (0.0)0 (0.0)GMFCS_IV0 (0.0)0 (0.0)0 (0.0)MACS, N (%)0.176MACS_I4 (25.0)1 (11.1)3 (42.9)MACS_II6 (37.5)3 (33.3)3 (42.9)MACS_III6 (37.5)5 (55.6)1 (14.3)MACS_IV0 (0.0)MRICS, N (%)0.504Type I1 (6.2)1 (11.1)0 (0.0)Type II5 (31.2)2 (22.2)3 (42.9)Type III10 (62.5)6 (66.7)4 (57.1)Type IV0 (0.0)0 (0.0)0 (0.0)Type V0 (0.0)0 (0.0)0 (0.0)

GMFCS, Gross Motor Function Classification System; MACS, Manual Ability Classification System; MRICS, MRI classification system. (*) Significant p-value (<0.05) of non-parametric comparisons between groups (Left vs. Right hemiplegia).

Representative examples of structural brain lesions in children with unilateral cerebral palsy, displayed on FLAIR and T1-weighted magnetic resonance images. Images illustrate variability in lesion location and extent across participants. The left hemisphere is shown on the right side of each image according to radiological convention.

AHA and BBT scores at T0, T1, T2, and T3 are summarized in Figures 2A–C. Primary linear mixed-effects models indicate that BBT_dom scores significantly improved at T1, T2, and T3 compared to T0; moreover, age had a significant positive effect, with each additional year of age associated with a score increase, while sex was not significant (Model 1.1). BBT_non_dom scores showed a significant increase at T2, with no significant effects of Age or Sex (Model 1.2). AHA scores increased significantly over time. Specifically, scores at T1 and T2 were significantly higher than at T0, while age and sex were not significantly associated with AHA scores (Model 1.3) (see Table 2).

Longitudinal changes in motor function outcomes across time points. Mean scores and standard deviations are shown for (A) Box and Block Test (BBT) of the dominant hand, (B) BBT of the non-dominant hand, and (C) Assisting Hand Assessment (AHA). T0: Baseline; T1: post-intervention; T2: 8 week follow-up; T3: 24 week follow-up. Error bars represent standard deviations.

ModelVariableCoefficientStd_errorT-valueP-valueModel 1. 1
Dependent variable:
BBT dominant hand scores(Intercept)29.6527.0634.1980.001TIMET14.11.6332.5110.016*TIMET25.6191.6443.4180.001**TIMET35.3161.6983.1310.003**Age2.130.5933.5930.002**Sex (Male)4.2074.1111.0230.325Model 1.2
Dependent variable:
BBT non-dominant hand scores(Intercept)10.6388.8321.2050.238TIMET11.9141.0551.8140.077TIMET22.8861.0752.6860.010*TIMET32.2381.1861.8880.065Age1.0570.7021.5070.139Sex (Male)2.2056.4690.3410.739Model 1.3.
Dependent variable: AHA scores(Intercept)44.5187.8895.643<0.001TIMET12.8950.8143.5590.001**TIMET22.3290.8482.7450.009**TIMET31.5890.9521.6690.101Age0.4240.6040.7030.485Sex (Male)2.0016.3280.3160.757

Primary linear mixed-effect models examining changes in clinical outcomes over time.

Separate models were run for each outcome variable: Assisting Hand Assessment (AHA) and Box and Block Test for non-dominant and dominant hand. All models were adjusted for time (T1, T2, T3), age, sex, and lesioned hemisphere (right vs. left). Asterisks indicate statistical significance (*p < 0.05, **p < 0.01).

Secondary models showed that the strongest associations emerged between motor outcomes (BBT_non_dom and AHA changes over time), and ipsilesional temporal, thalamus, brainstem, and corpus callosum (adjusted p-value < 0.01**). Additional associations were also observed with frontal and occipital lobe lesions (p < 0.05*). Instead, no region was associated with changes in the BBT_dominant. These results indicate that greater lesion involvement in these regions, as reflected by higher sqMRI scores, was associated with reduced response in the non-dominant hand to the AOT treatment over time. Age was positively associated with motor outcomes, while sex did not significantly influence outcomes. Table 3 reports coefficients, p-values, and FDR-corrected p-values for each brain region evaluated. Finally, models examining the effects of lesion type and timing (MRICS categories) on motor outcomes (BBT_dom, BBT_non_dom, and AHA) adjusted for TIME, Age, and Sex revealed no significant effect of MRICS variables (see Supplementary Table S2).

PredictorAHABBT_non dominantBBT dominantEstimateAdjusted p-valueEstimateAdjusted p-valueEstimateAdjusted p-valueIpsilesional frontal−8.7510.071−7.8350.041*−6.4060.116Ipsilesional temporal−7.9420.008**−7.1560.006**−1.5530.711Ipsilesional parietal−3.2470.396−4.1690.196−2.6870.668Ipsilesiona occipital−7.8910.031*−6.7340.023*−2.1030.668Ipsilesional_PLIC−13.0880.198−12.5410.1968.1440.668Ipsilesional_Lenticular−11.6720.155−8.4270.196−9.8090.183Ipsilesional_Thalamus−19.0470.008**−16.2190.007**−1.2950.975Ipsilesional_Brainstem−19.0470.008**−16.2190.007**−1.2950.975Corpus callosum−12.1480.008**−9.1780.013*−6.2610.116Contralesional frontal19.1600.15514.1280.1960.4440.975Contralesional temporal19.1600.15514.1280.1960.4440.975Contralesional parietal9.4540.1556.7970.196−0.1230.975Contralesiona occipital19.1600.15514.1280.1960.4440.975Cerebellum−9.8390.444−6.6300.536−13.5880.282

Secondary linear mixed-effects models examining association between structural brain lesions and clinical outcomes.

Separate models were run for each brain region (sqMRI score) and each outcome variable: Assisting Hand Assessment (AHA) and Box and Block Test for the dominant and non-dominant hands. All models were adjusted for time (T1, T2, T3), age, sex, and lesioned hemisphere (right vs. left). P-values were corrected for multiple comparisons using the False Discovery Rate (FDR) method. Asterisks indicate statistical significance after FDR correction (*p < 0.05, **p < 0.01). Regions with no observed lesions across the sample (Ipsilesional and Contralesional Caudate, Contralesional PLIC, Contralesional Lenticular nucleus, Contralesional Thalamus, and Contralesional Brainstem) were excluded from modeling.

4 Discussion

To our knowledge, the current study is the first to explore structural brain lesions potentially associated with AOT effects in the context of rehabilitation interventions in children with UCP. Our results demonstrate a significant increase in BBT scores for the dominant hand at all post-treatment time points, for the non-dominant hand