Accurate and reliable measurement of the severity of dystonia is essential for the indication, evaluation, monitoring and fine‐tuning of treatments. Assessment of dystonia in children and adolescents with dyskinetic cerebral palsy (CP) is now commonly performed by visual evaluation either directly in the doctor’s office or from video recordings using standardized scales. Both methods lack objectivity and require much time and effort of clinical experts. Only a snapshot of the severity of dyskinetic movements (i.e., choreoathetosis and dystonia) is captured, and they are known to fluctuate over time and can increase with fatigue, pain, stress or emotions, which likely happens in a clinical environment. The goal of this study was to investigate whether it is feasible to use home‐based measurements to assess and evaluate the severity of dystonia using smartphonecoupled inertial sensors and machine learning. Video and sensor data during both active and rest situations from 12 patients were collected outside a clinical setting. Three clinicians analyzed the videos and clinically scored the dystonia of the extremities on a 0–4 scale, following the definition of amplitude of the Dyskinesia Impairment Scale. The clinical scores and the sensor data were coupled to train different machine learning models using cross‐validation. The average F1 scores (0.67 ± 0.19 for lower extremities and 0.68 ± 0.14 for upper extremities) in independent test datasets indicate that it is possible to detected dystonia automatically using individually trained models. The predictions could complement standard dyskinetic CP measures by providing frequent, objective, real‐world assessments that could enhance clinical care. A generalized model, trained with data from other subjects, shows lower F1 scores (0.45 for lower extremities and 0.34 for upper extremities), likely due to a lack of training data and dissimilarities between subjects. However, the generalized model is reasonably able to distinguish between high and lower scores. Future research should focus on gathering more high‐quality data and study how the models perform over the whole day.

Introduction: Video-based clinical rating plays an important role in assessing dystonia and monitoring the effect of treatment in dyskinetic cerebral palsy (CP). However, evaluation by clinicians is time-consuming, and the quality of rating is dependent on experience. The aim of the current study is to provide a proof-of-concept for a machine learning approach to automatically assess scoring of dystonia using 2D stick figures extracted from videos. Model performance was compared to human performance. Methods: A total of 187 video sequences of 34 individuals with dyskinetic CP (8–23 years, all non-ambulatory) were filmed at rest during lying and supported sitting. Videos were scored by three raters according to the Dyskinesia Impairment Scale (DIS) for arm and leg dystonia (normalized scores ranging from 0–1). Coordinates in pixels of the left and right wrist, elbow, shoulder, hip, knee and ankle were extracted using DeepLabCut, an open source toolbox that builds on a pose estimation algorithm. Within a subset, tracking accuracy was assessed for a pretrained human model and for models trained with an increasing number of manually labeled frames. The mean absolute error (MAE) between DeepLabCut’s prediction of the position of body points and manual labels was calculated. Subsequently, movement and position features were calculated from extracted body point coordinates. These features were fed into a Random Forest Regressor to train a model to predict the clinical scores. The model performance trained with data from one rater evaluated by MAEs (model-rater) was compared to inter-rater accuracy. Results: A tracking accuracy of 4.5 pixels (approximately 1.5 cm) could be achieved by adding 15–20 manually labeled frames per video. The MAEs for the trained models ranged from 0.21 ± 0.15 for arm dystonia to 0.14 ± 0.10 for leg dystonia (normalized DIS scores). The inter-rater MAEs were 0.21 ± 0.22 and 0.16 ± 0.20, respectively. Conclusion: This proof-of-concept study shows the potential of using stick figures extracted from common videos in a machine learning approach to automatically assess dystonia. Sufficient tracking accuracy can be reached by manually adding labels within 15–20 frames per video. With a relatively small data set, it is possible to train a model that can automatically assess dystonia with a performance comparable to human scoring.

To maximize effects of dorsal leaf ankle foot orthoses (AFOs) on gait in people with bilateral plantarflexor weakness, the AFO properties should be matched to the individual. However, how AFO properties interact regarding their effect on gait function is unknown. We studied the interaction of AFO bending stiffness with neutral angle and footplate stiffness on the effect of bending stiffness on walking energy cost, gait kinematics and kinetics in people with plantarflexor weakness by employing predictive simulations. Our simulation framework consisted of a planar 11 degrees of freedom model, containing 11 muscles activated by a reflex-based neuromuscular controller. The controller was optimized by a comprehensive cost function, predominantly minimizing walking energy cost. The AFO bending and footplate stiffness were modelled as torsional springs around the ankle and metatarsal joint. The neutral angle of the AFO was defined as the angle in the sagittal plane at which the moment of the ankle torsional spring was zero. Simulations without AFO and with AFO for 9 bending stiffnesses (0–14 Nm/degree), 3 neutral angles (0–3-6 degrees dorsiflexion) and 3 footplate stiffnesses (0–0.5–2.0 Nm/degree) were performed. When changing neutral angle towards dorsiflexion, a higher AFO bending stiffness minimized energy cost of walking and normalized joint kinematics and kinetics. Footplate stiffness mainly affected MTP joint kinematics and kinetics, while no systematic and only marginal effects on energy cost were found. In conclusion, the interaction of the AFO bending stiffness and neutral angle in bilateral plantarflexor weakness, suggests that these should both be considered together when matching AFO properties to the individual patient.

Background: In this systematic review we investigate which instrumented measurements are available to assess motor impairments, related activity limitations and participation restrictions in children and young adults with dyskinetic cerebral palsy. We aim to classify these instrumented measurements using the categories of the international classification of functioning, disability and health for children and youth (ICF-CY) and provide an overview of the outcome parameters. Methods: A systematic literature search was performed in November 2019. We electronically searched Pubmed, Embase and Scopus databases. Search blocks included (a) cerebral palsy, (b) athetosis, dystonia and/or dyskinesia, (c) age 2-24 years and (d) instrumented measurements (using keywords such as biomechanics, sensors, smartphone, and robot). Results: Our search yielded 4537 articles. After inspection of titles and abstracts, a full text of 245 of those articles were included and assessed for further eligibility. A total of 49 articles met our inclusion criteria. A broad spectrum of instruments and technologies are used to assess motor function in dyskinetic cerebral palsy, with the majority using 3D motion capture and surface electromyography. Only for a small number of instruments methodological quality was assessed, with only one study showing an adequate assessment of test-retest reliability. The majority of studies was at ICF-CY function and structure level and assessed control of voluntary movement (29 of 49) mainly in the upper extremity, followed by assessment of involuntary movements (15 of 49), muscle tone/motor reflex (6 of 49), gait pattern (5 of 49) and muscle power (2 of 49). At ICF-CY level of activities and participation hand and arm use (9 of 49), fine hand use (5 of 49), lifting and carrying objects (3 of 49), maintaining a body position (2 of 49), walking (1 of 49) and moving around using equipment (1 of 49) was assessed. Only a few methods are potentially suitable outside the clinical environment (e.g. inertial sensors, accelerometers). Conclusion: Although the current review shows the potential of several instrumented methods to be used as objective outcome measures in dyskinetic cerebral palsy, their methodological quality is still unknown. Future development should focus on evaluating clinimetrics, including validating against clinical meaningfulness. New technological developments should aim for measurements that can be applied outside the laboratory.

Background: The stiffness of a dorsal leaf AFO that minimizes walking energy cost in people with plantarflexor weakness varies between individuals. Using predictive simulations, we studied the effects of plantarflexor weakness, passive plantarflexor stiffness, body mass, and walking speed on the optimal AFO stiffness for energy cost reduction. Methods: We employed a planar, nine degrees-of-freedom musculoskeletal model, in which for validation maximal strength of the plantar flexors was reduced by 80%. Walking simulations, driven by minimizing a comprehensive cost function of which energy cost was the main contributor, were generated using a reflex-based controller. Simulations of walking without and with an AFO with stiffnesses between 0.9 and 8.7 Nm/degree were generated. After validation against experimental data of 11 people with plantarflexor weakness using the Root-mean-square error (RMSE), we systematically changed plantarflexor weakness (range 40–90% weakness), passive plantarflexor stiffness (range: 20–200% of normal), body mass (+ 30%) and walking speed (range: 0.8–1.2 m/s) in our baseline model to evaluate their effect on the optimal AFO stiffness for energy cost minimization. Results: Our simulations had a RMSE < 2 for all lower limb joint kinetics and kinematics except the knee and hip power for walking without AFO. When systematically varying model parameters, more severe plantarflexor weakness, lower passive plantarflexor stiffness, higher body mass and walking speed increased the optimal AFO stiffness for energy cost minimization, with the largest effects for severity of plantarflexor weakness. Conclusions: Our forward simulations demonstrate that in individuals with bilateral plantarflexor the necessary AFO stiffness for walking energy cost minimization is largely affected by severity of plantarflexor weakness, while variation in walking speed, passive muscle stiffness and body mass influence the optimal stiffness to a lesser extent. That gait deviations without AFO are overestimated may have exaggerated the required support of the AFO to minimize walking energy cost. Future research should focus on improving predictive simulations in order to implement personalized predictions in usual care. Trial Registration Nederlands Trial Register 5170. Registration date: May 7th 2015. http://www.trialregister.nl/trialreg/admin/rctview.asp?TC=5170.

Neuromusculoskeletal models can be used to evaluate aberrant muscle function in cerebral palsy (CP), for example by estimating muscle and joint contact forces during gait. However, to be accurate, models should include representative musculotendon parameters. We aimed to estimate personalised parameters that capture the mechanical behaviour of the plantarflexors in children with CP and typically developing (TD) children. Ankle angle (using motion capture), torque (using a load-cell), and medial gastrocnemius fascicle lengths (using ultrasound) were measured during slow passive ankle dorsiflexion rotation for thirteen children with spastic CP and thirteen TD children. Per subject, the measured rotation was input to a scaled OpenSim model to simulate the torque and fascicle length output. Musculotendon model parameters were personalised by the best match between simulated and experimental torque–angle and fascicle length-angle curves according to a least-squares fit. Personalised tendon slack lengths were significantly longer and optimal fibre lengths significantly shorter in CP than model defaults and than in TD. Personalised tendon compliance was substantially higher in both groups compared to the model default. The presented method to personalise musculotendon parameters will likely yield more accurate simulations of subject-specific muscle mechanics, to help us understand the effects of altered musculotendon properties in CP.

Accurate predictive simulations of human gait rely on optimisation criteria to solve the system's redundancy. Defining such criteria is challenging, as the objectives driving the optimization of human gait are unclear. This study evaluated how minimising various physiologically-based criteria (i.e., cost of transport, muscle activity, head stability, foot–ground impact, and knee ligament use) affects the predicted gait, and developed and evaluated a combined, weighted cost function tuned to predict healthy gait. A generic planar musculoskeletal model with 18 Hill-type muscles was actuated using a reflex-based, parameterized controller. First, the criteria were applied into the base simulation framework separately. The gait pattern predicted by minimising each criterion was compared to experimental data of healthy gait using coefficients of determination (R²) and root mean square errors (RMSE) averaged over all biomechanical variables. Second, the optimal weighted combined cost function was created through stepwise addition of the criteria. Third, performance of the resulting combined cost function was evaluated by comparing the predicted gait to a simulation that was optimised solely to track experimental data. Optimising for each of the criteria separately showed their individual contribution to distinct aspects of gait (overall R²: 0.37–0.56; RMSE: 3.47–4.63 SD). An optimally weighted combined cost function provided improved overall agreement with experimental data (overall R²: 0.72; RMSE: 2.10 SD), and its performance was close to what is maximally achievable for the underlying simulation framework. This study showed how various optimisation criteria contribute to synthesising gait and that careful weighting of them is essential in predicting healthy gait.

Movement of skin markers with respect to their underlying bone (i.e. soft tissue artifacts (STAs)) might corrupt the accuracy of marker-based movement analyses. This study aims to quantify STAs in 3D for foot markers and their effect on multi-segment foot kinematics as calculated by the Oxford and Rizzoli Foot Models (OFM, RFM). Fifteen subjects with asymptomatic feet were seated on a custom-made loading device on a computed tomography (CT) table, with a combined OFM and RFM marker set on their right foot. One unloaded reference CT-scan with neutral foot position was performed, followed by 9 loaded CT-scans at different foot positions. The 3D-displacement (i.e. STA) of each marker in the underlying bone coordinate system between the reference scan and other scans was calculated. Subsequently, segment orientations and joint angles were calculated from the marker positions according to OFM and RFM definitions with and without STAs. The differences in degrees were defined as the errors caused by the marker displacements. Markers on the lateral malleolus and proximally on the posterior aspect of the calcaneus showed the largest STAs. The hindfoot-shank joint angle was most affected by STAs in the most extreme foot position (40° plantar flexion) in the sagittal plane for RFM (mean: 6.7°, max: 11.8°) and the transverse plane for OFM (mean: 3.9°, max: 6.8°). This study showed that STAs introduce clinically relevant errors in multi-segment foot kinematics. Moreover, it identified marker locations that are most affected by STAs, suggesting that their use within multi-segment foot models should be reconsidered.

Background: Children with cerebral palsy often show deviating calf muscle activation patterns during gait, with excess activation during early stance and insufficient activation during push-off. Research question: Can children with cerebral palsy improve their calf muscle activation patterns during gait using one session of biofeedback-driven gaming? Methods: Eighteen children (6–17 y) with spastic cerebral palsy received implicit game-based biofeedback on electromyographic activity of the calf muscle (soleus or gastrocnemius medialis) while walking on a treadmill during one session. Biofeedback alternately aimed to reduce early stance activity, increase push-off activity, and both combined. Early stance and push-off activity and the double-bump-index (early stance divided by push-off activity) were determined during baseline and walking with feedback. Changes were assessed at group level using repeated measures ANOVA with simple contrast or Friedman test with post-hoc Wilcoxon signed rank test, as well as individually using independent t-tests or Wilcoxon rank sum tests. Perceived competence and interest-enjoyment were assessed through a questionnaire. Results: Children successfully decreased their electromyographic activity during early stance feedback trials (relative decrease of 6.8 ± 12.2 %, P = 0.025), with a trend during the combined feedback trials (6.5 ± 13.9 %, P = 0.055), and increased their electromyographic activity during push-off feedback trials (8.1 ± 15.8 %, P = 0.038). Individual improvements were seen in twelve of eighteen participants. All children experienced high levels of interest-enjoyment (8.4/10) and perceived competence (8.1/10). Significance: This exploratory study suggests that children with cerebral palsy can achieve small within-session improvements of their calf muscle activation pattern when provided with implicit biofeedback-driven gaming in an enjoyable manner. Follow-up gait training studies can incorporate this method to assess retention and long-term functional benefits of electromyographic biofeedback-driven gaming.

Background: Spasticity, i.e. stretch hyperreflexia, increases joint resistance similar to symptoms like hypertonia and contractures. Botulinum neurotoxin-A (BoNT-A) injections are a widely used intervention to reduce spasticity. BoNT-A effects on spasticity are poorly understood, because clinical measures, e.g. modified Ashworth scale (MAS), cannot differentiate between the symptoms affecting joint resistance. This paper distinguishes the contributions of the reflexive and intrinsic pathways to ankle joint hyper-resistance for participants treated with BoNT-A injections. We hypothesized that the overall joint resistance and reflexive contribution decrease 6 weeks after injection, while returning close to baseline after 12 weeks. Methods: Nine participants with spasticity after spinal cord injury or after stroke were evaluated across three sessions: 0, 6 and 12 weeks after BoNT-A injection in the calf muscles. Evaluation included clinical measures (MAS, Tardieu Scale) and motorized instrumented assessment using the instrumented spasticity test (SPAT) and parallel-cascade (PC) system identification. Assessments included measures for: (1) overall resistance from MAS and fast velocity SPAT; (2) reflexive resistance contribution from Tardieu Scale, difference between fast and slow velocity SPAT and PC reflexive gain; and (3) intrinsic resistance contribution from slow velocity SPAT and PC intrinsic stiffness/damping. Results: Individually, the hypothesized BoNT-A effect, the combination of a reduced resistance (week 6) and return towards baseline (week 12), was observed in the MAS (5 participants), fast velocity SPAT (2 participants), Tardieu Scale (2 participants), SPAT (1 participant) and reflexive gain (4 participants). On group-level, the hypothesis was only confirmed for the MAS, which showed a significant resistance reduction at week 6. All instrumented measures were strongly correlated when quantifying the same resistance contribution. Conclusion: At group-level, the expected joint resistance reduction due to BoNT-A injections was only observed in the MAS (overall resistance). This observed reduction could not be attributed to an unambiguous group-level reduction of the reflexive resistance contribution, as no instrumented measure confirmed the hypothesis. Validity of the instrumented measures was supported through a strong association between different assessment methods. Therefore, further quantification of the individual contributions to joint resistance changes using instrumented measures across a large sample size are essential to understand the heterogeneous response to BoNT-A injections.