Paper Submission & Registration
8th Dutch Bio-Medical Engineering Conference
Go-previous
12:30   Rehabilitation - II
Chair: Kenneth Meijer
12:30
15 mins
Touch - A new dimension in rehabilitation robotics
Elisabeth Wilhelm
Abstract: Robots are widely accepted tools in the context of neurorehabilitation. They can assist the therapist in heavy load bearing tasks and provide high intensity training by increasing the amount of repetitions. Since the robot is steered by a computational unit it is also easy to combine robotic rehabilitation with gamification to increase the motivation of the patient. In accordance with recent findings about the contribution of neuroplasticity in neurorehabilitation, a lot of effort has been spent to allow training of functional tasks. Furthermore, adaptive control strategies have been implemented to enforce active participation of the patient (Gassert et al. 2018). However, the sensory feedback given by the robotic devices is still limited. Most devices only provide auditory and visual feedback. Furthermore, rendering of the interaction forces has been considered to provide force-feedback for example in form of virtual walls (Nef et al. 2016). Specific robotic devices for providing vestibular stimulation have been used in sleep research (van Sluijs et al. 2020). Tactile stimulation if provided at all is limited to relatively small skin areas such as one fingertip (Lambercy et al. 2011, Wilhelm et al. 2016). This is especially crucial considering the wide variety of neurologic conditions that are linked to altered tactile sensation such as for example Parkinson’s disease (Nolano et al. 2008). Patients with tactile deficits could benefit from sensory training (Carey et al. 2011). The biggest challenge in the development of a robotic device that can provide sensory training is that tactile sensation has a resolution in the micro-meter range. At the same time the organ that perceives this stimulation is spread across the whole body. To address this, we combine sophisticated micro- and macro engineering techniques to develop robots for sensory training.
12:45
15 mins
Reliable and valid quantification of upper limb impairment with a robotic device in stroke patients and adults with cerebral palsy
Levinia van der Velden, Joyce Benner, Bram Onneweer, Claudia Haarman, Ruud Selles, Gerard Ribbers, Marij Roebroeck
Abstract: Assessment of upper limb impairment in patients with upper motor neuron lesion should ideally be rater-independent, reliable, and valid. In this study, we investigated the test-retest reliability and construct validity of the assessment of muscle weakness, abnormal synergy, spasticity, and changes in viscoelastic joint properties of the elbow using a single robotic device named the Shoulder-Elbow-Perturbator (SEP). In addition, we evaluated the criterion validity of the spasticity and abnormal synergy assessments. Test-retest reliability was evaluated for all four impairments in 9 chronic stroke patients, 20 adults with cerebral palsy, and 25 healthy controls. For each impairment, the intraclass correlation coefficients (ICC), standard error of measurement (SEM), and smallest detectable change (SDC) were calculated. Construct validity was tested for all impairments by evaluating group differences between healthy controls and patients, and between both patient groups. For abnormal synergy, criterion validity was checked by correlation analyses with the Upper Limb Fugl-Meyer Assessment (stroke patients) and Test of Arm Selective Control (adults with cerebral palsy); for spasticity, the Modified Tardieu Scale was assessed in all patients. The results showed an excellent ICC for all parameters (ICC2,1>0.75). Group differences between patients and healthy controls were found for all upper limb parameters (p<0.05), except for abnormal synergy (p=0.38). In addition, there were no group differences between healthy controls and adults with cerebral palsy for abnormal synergy (p=1,0), nor for extension muscle strength between healthy controls and stroke patients (p= 0.23). Correlation for abnormal synergy in stroke patients was good (r=-0.69, p=0.04), but poor for adults with cerebral palsy (r=-0.30, p=0.19). Modified Tardieu Scale correlated well with the robotic device in all patients (r>0.66, p<0.05). Our findings show that upper limb impairments can be quantified reliably and validly by the Shoulder-Elbow-Perturbator. Moreover, using a single device ensures more rater-independency in assessing upper limb impairment in daily practice than current clinical measures.
13:00
15 mins
Toward patient-specific neuro-mechanical analysis via wearable sensorized garment and rapid musculoskeletal modeling
Donatella Simonetti, Bart F. J. M Koopman, Massimo Sartori
Abstract: Nowadays, clinical diagnostic tools for post-stroke motor deficits are based on rapid and subjective metrics: the functional ambulation categories (FACs) [1]. On the other hand, greater accuracy is provided by well-equipped biomechanical laboratories in combination with electromyography (EMG)-driven musculoskeletal modeling and simulation [2] i.e. providing a quantitative assessment of the patient’s neuro-mechanical actual state. However, the process of muscle localization for electrode placement and joint angle measurement is not always viable in clinical environments where clinicians need to rapidly and objectively evaluate patient-specific neuro-physical conditions over time. We are aiming to develop an advanced technology that combines (1) a fully wearable soft sensorized garment and (2) an automatic algorithm for muscles localization to reduce the set-up time and prevent human error, and finally drive (3) a real-time framework for accurate patient-specific neuro-mechanical modeling. A healthy subject was equipped with a flexible garment with an integrated grid of 64 equally distributed EMG electrodes around the lower leg and 35 reflective markers. The 64-electrode space was reduced in 5 muscle-specific clusters applying non-negative matrix factorization (NNMF) [3]during slow locomotion at 1km/h. Afterwards, we extracted 5 average muscle activations during locomotion at different speeds 1, 3, and 5 km/h, and used them to estimate (2) torque at the ankle joint using EMG-driven musculoskeletal model. The 64 EMG channels were reduced in 5 synergistic active regions. The extracted averaged activations during each locomotion speed resembled with good accuracy the activation from bipolar electrodes (𝑅2=0,95±0.01; RMSE = 0,027 ±0,008 ). Afterwards, the musculoskeletal model driven by the automatically extracted muscle-specific activation reproduced experimental ankle torques during gait at different speeds. We obtained promising preliminary results of a new approach for lower leg muscle localization using NNMF-based extraction of muscle activations during locomotion and ankle torque estimation using multi-channel EMG-driven musculoskeletal modeling. The combination of a flexible sensorized garment and the automatic procedure of muscle activity extraction added to the framework for neuromuscular modeling has potential to become a resource for clinicians. It will provide rapidity and an enhanced perspective on the patient’s musculoskeletal system helping to tailor a subject-specific rehabilitation treatment for optimal recovery.
13:15
15 mins
Why are clinicians still using the modified ashworth scale? A systematic review of the diagnostic levels of evidence of diagnostic robotic devices for measuring viscoelastic joint properties and spasticity
Maaike de Koff
Abstract: Many different diagnostic robotic devices have been developed in the last decades to quantify viscoelastic properties and spasticity of patients with upper motor neuron lesions. However, in clinical practice, subjective and nonvalid clinical scales such as the Modified Ashworth Scale are still commonly used. We performed a systematic review to investigate one possible reason of disuse of robotic devices in clinical practice, by evaluating the diagnostic level of evidence of robotic device studies to assess viscoelastic properties and spasticity. We performed a literature search. Inclusion criteria were: children and adults with stroke or cerebral palsy, use of a robotic device to measure viscoelastic properties and/or spasticity of limbs. Two of the authors independently screened all articles. To determine the diagnostic level of evidence, a classification was made following Sackett et al.[1], identifying six different phases. These phases were based on the number of participants and the design of the study to determine the validity, diagnostic accuracy and the added value for treatment effects. The initial search yielded 1697 articles; after screening, 37 articles were included. Most studies measured the upper limb (70%) in stroke patients (76%). Most studies (46%) had a diagnostic level of evidence of phase 0, meaning these studies included less than ten patients. The highest level of evidence we found was phase 2a (43%), these studies correlated the test values of the robotic device with a clinical test or within subgroups. The remaining studies (11%) had a level of evidence of phase 1, determining the range of values of the robotic test. None of the studies had a phase 2b, 3 or 4, determining the diagnostic accuracy or added value for treatment effects. In conclusion, most studies included too few participants for establishing clinical evidence. In addition, none of the studies tested their diagnostic robotic device for accuracy (phase 2b) or added value (> phase 3). This review shows that the type of evidence needed for implementing robotic devices in clinical practice is currently lacking. Further research should focus more one providing this evidence, to get clinicians using the robotic devices instead of the Modified Ashworth Scale. 1Sackett DL, Haynes RB. The architecture of diagnostic research. BMJ. 2002;324(7336):539-541.


end %-->