Instrumented dystonia and choreoathetosis assessment protocol (IDCA) of upper limb movements in cerebral palsy

The general aim of this doctoral project was to increase our insights into the qualitative and quantitative assessment methods for childhood-onset dystonia and choreoathetosis. Dystonia and choreoathetosis are movement disorders that interfere with activities of daily living and have a tremendous impact on quality of life. Dystonia and choreoathetosis can occur in children after a brain lesion near birth, as a consequence of a genetic defect, or without known origin, but its assessment is complicated due to the multi-faceted presentation of both movement disorders, especially if they present concurrently. For the qualitative part of this project, we focused on the assessment of dystonia and choreoathetosis using the dyskinesia impairment scale (DIS). This scale was specifically developed to assess dystonia and choreoathetosis in individuals with dyskinetic cerebral palsy. We expanded our focus towards individuals with inherited or idiopathic dystonias, since there is currently no assessment tool available that evaluates both individuals with primary and secondary dystonia. For the quantitative part of this project, we focused on upper limb movements, since many individuals with dyskinetic cerebral palsy can perform simple upper limb tasks. We implemented an instrumented assessment protocol using different methodologies during functional upper limb tasks in individuals with dyskinetic cerebral palsy.

In chapter one, we explored the reliability and validity of the DIS in individuals with inherited and/or idiopathic dystonia to evaluate whether the DIS can be used to asses presence and severity of dystonia and choreoathetosis in this group. We found excellent inter-rater reliability for the total DIS scores and the dystonia subscale amongst three raters, indicating that presence and severity of dystonia can be reliably assessed in individuals with inherited or idiopathic dystonia. Inter-rater reliability for the presence and severity of choreoathetosis was somewhat lower, but still good. The lower inter-rater reliability for choreoathetosis could be attributed to lower presence of choreoathetosis in our study sample, and thus a higher presence of ‘zeros’, signifying absence of choreoathetosis. All scores showed high test-retest reliability, indicating that the DIS is a reliable scale to assess presence and severity of dystonia and/or choreoathetosis over time in a study sample with primary dystonia. This is important if we wish to use this scale to assess the effect of interventions. We found a minimal detectable difference of 11%, which means that if we wish use the DIS to detect an intervention effect, this effect should be higher than 11% of the total score on the DIS. These findings show that the DIS is a promising tool to assess dystonia and choreoathetosis in individuals diagnosed with inherited or idiopathic dystonia.

The aim of chapter two was to develop a shorter, more user-friendly version of the DIS, the DIS-II. To attain this objective, we merged the DIS scores from chapter one with previously collected scores in studies for the DIS development. This combination allowed us to work with a larger set of scores, which is a requirement for the implementation of the Rasch measurement model analysis. Subsequently, we invited clinicians (n=31) from five European countries who had worked with the DIS to a full-day expert meeting to provide input on the included tasks, number of body regions and rating scale steps. Consensus was reached through iterative discussion with the research group, resulting in a proposal for the DIS-II for later evaluation with Rasch analysis. Finally, we used the Rasch measurement model analysis to assess the construct validity and scale reliability of the newly developed scale. Construct validity consisted of rating scale functioning, fit statistics, point-measure correlations and targeting, whereas scale reliability included item and person reliability. Best reliability and construct validity were obtained when using 11 body regions (trunk was excluded) and scoring scale steps from 0 to 2 for rest and from 0 to 3 for action with 3= ‘action not performed’. The DIS-II requires 14 videos to record (instead of 26 for the DIS) and 88 items to score (instead of 144 in the DIS). Rating scale functioning and model fit were good, showing only two misfitting items for the DIS-II in the eye region. These items should thus be interpreted with caution. This is in agreement with the input of the consensus meeting that the eyes are the hardest body region to score, since eye movements often happen unconsciously. The DIS-II showed high internal consistency values and good separation, with individuals being divided into eight severity levels. The DIS-II thus provides a shorter, more user-friendly scale to assess dystonia and choreoathetosis. Additionally, the conversion of percentage scores to interval scores using Rasch conversions can enhance comparison of DIS-II scores over time and between individuals, regardless of the number of actions performed.

In chapter three, we focused on upper limb tasks using three-dimensional motion analysis. The movement pattern of individuals with dyskinetic cerebral palsy (CP) is characterized by increased variability due to the presence of dystonia and choreoathetosis, but few information on the psychometric properties of upper limb movements in dyskinetic CP is available. Therefore, we explored the psychometric properties of upper limb kinematics in individuals with dyskinetic CP during the execution of functional tasks. We aimed to explore i) within-session repeatability, ii) variability between individuals with dyskinetic CP and typically developing participants, iii) between-session repeatability and iv) differences in upper limb kinematics between individuals with and without dyskinetic CP. All participants were evaluated using three-dimensional motion analysis during the execution of a reach forward, reach and grasp vertical and reach sideways task and joint angles during task execution were obtained for the trunk, shoulder, elbow and wrist, as well as maximal velocity and trajectory deviation (spatio-temporal parameters). From the joint angles over time, we focused on the joint angle at point of task achievement, i.e. when the participants reach the objective to touch/grasp. All parameters – joint angle at point achievement and spatio-temporal parameters – were calculated for 2,4,6,8 and 10 repetitions and the change in measurement error per number of repetitions was evaluated. We found the lowest change – and thus most stable representation of the movement pattern – after 8 repetitions. This finding indicates that including a higher number of repetitions when evaluating individuals with dyskinetic CP is important to capture their variable movement pattern. Additionally, we found a significantly higher standard deviation for the majority of the joint angles and all spatio-temporal parameters for the dyskinetic CP group, confirming their higher movement variability. We found high between-session repeatability for most of the joint angles, with the exception of elbow pro/supination and scapula angles. For these joints, correct palpation of anatomical landmarks is crucial for their accuracy and should be carefully performed to assure good repeatability over time. Finally, we found higher wrist and elbow flexion, shoulder external rotation and trunk axial rotation at point of task achievement for the dyskinetic CP group, as well as higher trajectory deviation and lower maximal velocity. These findings confirm that individuals with dyskinetic CP have a variable movement pattern, characterized by deviant joint angles in comparison with typically developing peers, but less stereotyped in comparison with individuals with spastic CP.

In chapter four, we used the upper limb joint angles over time from chapter three, expanded with data collected in the Netherlands. The objectives of this study were two-fold: First, we explored whether individuals with dyskinetic CP showed higher intra-individual variability of upper limb joint angles over time in comparison with typically developing peers using statistical parametric mapping (SPM). Second, we used non-linear registration SPM to explore time and amplitude differences in the angular waveforms between both groups. SPM identifies significant clusters in a movement cycle, allowing to detect time-dependent differences between upper limb kinematics of two populations. With respect to the first objective, we found that individuals with dyskinetic CP show higher intra-individual variability at the level of all joints, but the occurrence of significant clusters was time-dependent and varied between tasks. At the level of the wrist, intra-individual variability was higher for the dyskinetic CP group for the full reaching cycle during forwards and sideways reaching, and for the first half of the reaching cycle during reach and grasp vertical. At the level of the trunk and shoulder, intra-individual variability was higher for the majority of or the full reaching cycle during all tasks. With respect to the second objective, we found higher mean wrist and elbow flexion for the dyskinetic CP group during all tasks for varying parts of the reaching cycle. These results revealed important information at specific joint levels and specific timings during the movement cycle which can contribute to individualized treatment strategies and assessment of their effect in individuals with dyskinetic CP.

In chapter five, we focused on wearable sensors, and more specifically inertial measurement units (IMUs) for the characterization of movement disorders in individuals with a neurological condition. We searched the literature to map sensor set-up, number of sensors, sensor location, included tasks, sensor features extracted from the acceleration and/or angular velocity signal, and significance of the extracted sensor parameters in multiple pathological populations. We identified 101 articles that matched our inclusion criteria, of which 56 researched Parkinson’s Disease (PD). Wrist(s), hand(s) and index finger(s) were the most popular sensor locations. Most frequently included tasks were: finger tapping, wrist pro/supination, keeping the arms extended in front of the body and finger-to-nose. This task inclusion can be explained from the presence of these tasks in the clinical scale to evaluate the symptoms of PD, the Unified Parkinson Disease Rating Scale (or the Movement Disorder Society Revised version). Most frequently calculated sensor features were mean, standard deviation, root-mean-square, ranges, skewness, kurtosis/entropy of acceleration and/or angular velocity, in combination with dominant frequencies/power of acceleration signals. The derived sensor features were dependent of the movement disorder under investigation. Amplitude-related parameters (mean and range of amplitude and amplitude decrement) were used in PD studies and are hypothesized to correlate with hypokinesia (reduction in movement amplitude). Velocity decrement and peak-to-peak, magnitude, and mean of angular velocity were additionally only used in PD studies and are hypothesized to relate to bradykinesia (slowing of movement). Finally, frequency-related parameters were mostly used to characterize tremor in PD and essential tremor. In conclusion, insights from PD studies can accelerate the development of wearable sensors protocols in the remaining pathologies, provided that there is sufficient attention for the standardisation of protocols, tasks, feasibility and data analysis methods.

In chapter six, we used the insights gained in chapter five to evaluate the reliability and discriminative ability of sensor features in individuals with and without dyskinetic CP. We placed IMUs on the upper arm, forearm and hand and collected acceleration and angular velocity signals during the execution of a reach forward, reach and grasp vertical and reach sideways task. We evaluated within-and between session reliability of maximal jerk and angular jerk, mean, maximal, root-mean square and sample entropy of acceleration and angular velocity, as well as the differences between the TD and dyskinetic CP group for all features. We additionally evaluated the reliability for all sensor features for long signals (the execution of eight reach forward repetitions) and short signals (one repetition, from hand on the knee to point of task achievement). We found highest reliability for the sensor features based on the long signals, for all features except for sample entropy (within-sessions) and jerk and sample entropy (between sessions). Jerk, angular jerk and maximal acceleration and angular velocity were significantly higher for the dyskinetic CP group in comparison with TD peers. These findings show that wearable sensors are a promising method to detect pathological movement patterns in individuals with movement disorders, providing attractive possibilities for instrumented assessments in a natural environment such as home or school. This can enhance adherence for long-term assessment and increase our insights into the fluctuating patterns depending on stress and arousal levels in individuals with dyskinetic CP.