Projects
Our projects aim at enabling place independent and motivating rehabilitation therapy thoughout the continuum of care
Smart Assessments
Personalized Training
Brain & Reward
Taste & Nutrition
DELTA
Leveraging Deep Learning to Assess Upper Limb Kinematics after Stroke with Off-the-shelf Webcams
The main goal of this study is to find a low-cost and low-complexity sensor setup which allows reliable estimation of upper body kinematics for stroke patients. This sensor configuration should be accessible to everybody and enable the monitoring of upper limb performance outside a motion laboratory. In the study, we will replicate typical clinical tests used in neurorehabilitation, such as the drinking task or box and block, in around 60 stroke patients. We collect movement and muscle activation from various systems (IMUs, EMGs, webcams, depth cameras, and 3D optical motion capture) and clinical data. Recorded data will be used to identify a minimal sensing configuration to reproduce kinematics and clinical scores accurately. Furthermore, the data will provide greater insight into the kinematic consequences of stroke and compensatory movement strategies. Data collection is conducted primarily at Balgrist Campus (ETH Zürich) and partly in Vitznau (cereneo/CEFIR motion lab).
A collaboration between ETHZ, ZHAW, UZH, USZ and CEFIR
Hand Dexterity
Novel technology-based methods to assess hand dexterity in real-life clinical settings
This project aims to develop and validate a biomechanical assessment and therapy platform for dexterity and finger individuation in a real-life clinical setting. This will help better understand the complex relationship between sensory and cognitive deficits of fine motor control in upper-extremity rehabilitation. State-of-the-art rehabilitation technology, assessing proprioception and finger individuation, will be tested for usability and feasibility in a clinical setting using a heterogenous neurological patient group. Further, during recovery, the relationship between these tools and clinically established dexterity measures will be analyzed. Findings will directly flow back to the adaptation of the analyzed technologies, which will lead towards minimally supervised training that enables patients’ self-assessment. This research will be exceptionally value-generating for patients and clinicians in neurorehabilitation, leading towards a better understanding of upper limb function, impairment, and recovery with the potential to help clinical decision-making for a more personalized and holistic therapy approach.
Activity Monitoring
Monitoring daily activities with wearable sensors
During their daily clinical life, patients undergo several therapies every day. While the methods and exercises used in these sessions are documented manually by the health care provider, the physical activity load of the patients remains only roughly quantified – a “black box” of in-patient rehabilitation. But overall daily physical activity plays a significant role in the rehabilitation, not only concerning movement ability but also regarding the cardiovascular system and cognitive function. With the rise of IMU Movement Sensor Technologies, we have a tool to look into this “black box”. This project aims to use IMU sensors to monitor stroke patients during their daily clinical routine and develop an algorithm to determine their daily physical activity load. However, patients spend only around 20% of their day in guided therapy. Therefore, in a second step, we would like to use the same sensors and algorithms to investigate and document the activity load of patients outside the therapy and gain insight into how these activities influence their rehabilitation journey.
EMG coherence analysis
EMG coherence analysis as a proof-of-concept for supra-spinal contribution in the gait cycle
During the last decades, science hasn’t reached a unanimous conclusion regarding the effectiveness of coherence analysis. Invasive animal experimentation based on spinal cord sectioning showed supra-spinal contribution to tibialis anterior control during the gait cycle. But in humans, such research on corticospinal control can’t be conducted for ethical and legal reasons. This indicates that muscular coherence analysis could work as a non-invasive predictor. Thus, this study will run a coherence analysis as a proof-of-concept based on a well-understood neuromuscular topic. Two electrodes on the tibialis anterior muscle gather data for intramuscular coherence, while a third electrode placed on the outer gastrocnemius muscle serves for intermuscular coherence analysis. Data is cleaned and labeled with Vicon, then processed in Matlab with the btk and neurospec22 libraries. Positive results will support coherence analysis as a suitable tool for further research. Ultimately, an automated program could directly plot muscular coherence with the recorded EMG signals, resulting in a non-invasive predictive biomarker.
StimuLoop - Handshake
Novel AI-based approach to quantify movement quality in gait
Current objectives of neurorehabilitation for gait mainly target functional outcomes, such as gait speed, while placing less emphasis on metrics to measure gait quality. One underlying contributor is that selecting movement parameters for gait quality treatment is time-consuming and complex. Physicians commonly use visual-based approaches to assess gait quality, even though the sensitivity of quality metrics depends on the assessor’s experience. An alternative approach is to use quantitative methods that rely on complicated instrumentation to translate into clinical terminology and treatment recommendations. We propose an AI-based clinical decision support system that merges the strengths of these two systems: exact outcomes of 3D motion capture are combined with target parameter recommendations given by clinical experts. We evaluate our automated clinical decision support workflow against purely visual-based recommendations or heuristic outputs of current analysis systems.
StimuLoop - Hyper Personalized Feedback
Personalized augmented feedback for post-stroke gait rehabilitation
Loss of mobility is a major deficit for stroke patients. It leads to a reduced autonomy and affects quality of life. Consequently, the main goal of stroke rehabilitation is to recover mobility and motor functions. In gait therapy, technical assistance, more specifically augmented feedback (FB), can help post-stroke patients to relearn correct walking patterns. It can be used in clinical environment with a therapist but also individually at home. Yet, there are no guidelines for selecting either FB modality (e.g., haptic, visual or audio) or FB signal characteristics (e.g., continuous or sequential). In this project, we investigate whether the correction of the gait pattern differs depending on the augmented FB features. This project is divided in two phases: a pilot study with healthy participants, and a study with post-stroke individuals. Gait adaptation, usability and acceptance are evaluated with different FB signals. This project paves the way towards a better understanding of augmented FB mechanisms and towards personalized feedback therapy.
Pertubation-based balance training
Perturbation-based balance training as a fall prevention strategy
All people hope to enjoy the best health in old age, free from impairment. This starkly contrasts with the 37 million fall injuries and 684’000 deaths from falling each year. Next to the acute danger of falling per se, just the fear of falling reduces mobility behaviour and negatively affects participation, socialisation, and quality of life. A frequently encountered downward spiral entails fear of falling -> reduced mobility behaviour -> accelerated neuromotor degeneration -> increased fall risk -> increased fear of falling. Over the past decades, research on how to alleviate this chain of events has converged on preventive balance training solutions. Perturbation-based balance training (PBT) mimics the accidental and unpredictable nature of slips and trips in daily life in a safe and controlled environment. PBT allows individuals to practice rapid and effective reactive gait adjustments as a fall prevention strategy and acts as a confidence booster for secure and stable ambulation. The PBT study aims to evaluate the acute and midterm effects and the underlying mechanisms of a single training session in older adults.
Target Memory Reactivation
Target Memory activation through music
Rehabilitation success after a stroke depends not only on active training and rehearsal during the day but also on active memory consolidation processes facilitated by sleep. Recently, techniques to modulate memory consolidation in sleep via external stimulation have been developed. In targeted memory reactivation (TMR), participants learn to associate a particular cue (e.g., background melody or sound) with target information (e.g., word pairs, object locations, or a motor sequence) when awake. Many studies showed that the specific memory for the target information is facilitated when participants are re-exposed to the learned cues during consecutive nonrapid eye movement (NREM) sleep. TMR has been shown to benefit different kinds of learning, suggesting its potential beneficial role in rehabilitation. In this study, we investigate whether an acoustic TMR implementation in a home-based EEG system can enhance learning a modified gait pattern. The novel aspects of our study are (1) a novel TMR algorithm implemented in a home-based EEG system with automated NREM sleep detection and (2) a gamified trajectory learning task for the lower limb that requires continuous modification of the ankle trajectory .Data collection is conducted in the CAREN in Vitznau (cereneo/CEFIR motion lab). A collaboration between UZH, USZ and CEFIR.
Wanting and Liking
AI based detection of food wanting and liking from a webcam video
Food preference is a combination of food liking and food wanting, and although these act together to determine dietary behaviour, liking and wanting can be differentiated. Food liking can be assessed by facial recognition analysis. However, there is currently no objective method to assess food wanting. The global lack of objective assessment for this hinders our understanding of negative food-related behaviours. In this project, we will collect liking and wanting rating score data together with video to assess if it is feasible to use computer vision to predict both aspects of food preference. During the experiment, we will present participants with several breakfast items in the morning and ask them to rate how much they want or like this food item. A web camera is applied to capture facial expression changes while several physiological signals (such as heart rate and pupil diameter) are collected synchronously. In the end, an AI-based (Long Short-Term Memory neural network) classifier is trained based on video data and physiological signals to classify whether the participant wants this food.
WESLA
Wearable Sensor for Speech, Language and Voice analysis
In this project, we will create a wearable sensor system for health care provider and researchers which allows precise context-independent measurement of voice, speech as well as language parameters. The system incorporates the newest technological advances in capturing skin vibration signals from which a wide range of information on voice, speech and language performance as well as information related to coughing and swallowing can be extracted. Monitoring real-life performance will increase therapists knowledge on neurological patient’s ability to transfer therapy content into real-life and advance the field of neurorehabilitation. A wide range research areas beyond neurorehabilitation research would also strongly benefit of being able to capture intelligible speech in natural conversations.
NMDS
Neuropsychological Minimal Dataset after Stroke
We aim at harmonizing neuropsychological tests for the assessment of cognitive impairments after stroke. The goal is to determine a multilingual experts’ consensus-based, neuropsychological minimal dataset, harmonized across all countries of the European Economic Area. Current practice and the opinion of experts in neuropsychology is inquired in 2 surveys and the outcome is discussed in a round table meeting. This work forms the basis for further projects that are intended to drive digitisation.
A collaboration between UZH, Clinic Valens and CEFIR
Personalised Palates
Exploration of taste and smell rehabiliation
The aim of this project is to explore the potential of taste and smell rehabilitation in patients who have suffered a stroke. Neurodegenerative/neurological diseases (ND) can lead to loss of taste and smell. Our ability to taste and smell can not only provide gastronomical enjoyment but also elicit autobiographical memories, familiarity, aid social interaction, and contribute to cross-modal perception. Such loss can be detrimental to quality of life and lead to malnutrition. Sensory rehabilitation has been demonstrated to be effective at re-establishing sense of smell in some ND, for example, in patients with Parkinson’s Disease, but effectiveness in stroke patients is unknown, yet similar deficiencies are seen. Sensory, phenotypic and genotypic data are being gathered in healthy and ND affected populations to investigate differences in taste and smell ability. Following this, sensory rehabilitation trials will be implemented. Our preliminary results have shown a significant difference in taste and smell ability between those with stroke and healthy controls. Future analysis aims to assess taste and smell ability changes due to a sensory rehabilitation programme, also based on genetic predisposition to deficits in taste and smell.
A collaboration between CEFIR, LLI, and The University of Trieste.
From Palate to Plate
Exploration of the sensory palate and dietary intake
The aim of this project is to develop an algorithm that allows us to define what an individual’s food preferences are in order to create personalised meals that are healthy yet tasty. This will not only benefit healthy individuals but is also useful for patients suffering from neurological disorders, who have perceptual alterations in taste and smell. In this situation, patients tend to eat less or choose more palatable and unhealthy foods, eventually leading to malnutrition and to a further deterioration of health. Using standardised tests, we assess the two senses, specifically the ability to perceive and discriminate tastes and smells, investigating, also, participants’ food preferences by questionnaires. With this information, we were able to develop a ‘beta’ version of the algorithm. We are currently collecting phenotype and genotype data to extend and improve this. The next step of the project will be the implementation of the genetic data, within the already developed algorithm to obtain an even more precise sensory profile for the participants. Genetics plays a key role in taste perception; several polymorphisms are reported in literature to be associated with different taste sensitivities. The results of this project have the potential to improve dietary behaviour and general enjoyment of eating.
A collaboration between CEFIR, LLI, and The University of Trieste.
Recent Publications
Pohl J, Ryser A, Veerbeek JM, Verheyden G, Vogt J, Luft AR, Easthope CA. Classification of functional and non-functional arm use by inertial measurement units in subjects with upper limb impairment after stroke. Front Physiol 13, 952757 (2022)
Pohl J, Ryser A, Veerbeek J, Verheyden G, Vogt J, Luft AR, Easthope CA. Accuracy of gait and posture classification using movement sensors in individuals with mobility impairment after stroke. Front Physiol 13, 933987 (2022)
Zipser-Mohammadzada, F., Conway BA, Halliday DM, Zipser CM, Easthope CA, Curt A, Schubert M. Intramuscular coherence during challenging walking in incomplete spinal cord injury: Reduced high-frequency coherence reflects impaired supra-spinal control. Front Hum Neurosci 16, 927704 (2022)
Clément-Guillotin, C., Colombel, F., Easthope, C. A. & Fontayne, P. Being incidentally exposed to a sport context: same consequences on gender schema activation as being in a sport context? Mov Sport Sci – Sci Mot 61–72 (2022)
Mohammadzada F, Zipser CM, Easthope CA, Halliday DM, Conway BA, Curt A, Schubert M. Mind your step: Target walking task reveals gait disturbance in individuals with incomplete spinal cord injury. J Neuroeng Rehabil 19, 36 (2022)
Werner C, Schönhammer JG, Lambercy O, Luft AR, Demkó L., Easthope, CA. Using Wearable Inertial Sensors to Estimate Clinical Scores of Upper Limb Movement Quality in Stroke. Front Physiol 13, 877563 (2022)
Branscheidt M, Ejaz N, Xu J, Widmer M, Harran MD, Cortés JC, Kitago T, Celnik P, Hernandez-Castillo C, Diedrichsen J, Luft A, Krakauer JW. No evidence for motor-recovery-related cortical connectivity changes after stroke using resting-state fMRI. J Neurophysiol 127, 637–650 (2022).
Klamroth-Marganska, V., Giovanoli, S., Easthope, C.A., Schönhammer, J.G. Telerehabilitation Technology. 563–594 (2022) https://doi.org/10.1007/978-3-031-08995-4_25 In: Reinkensmeyer, D.J., Marchal-Crespo, L., Dietz, V. (eds) Neurorehabilitation Technology. Springer, Cham.
Werner C, Awai Easthope C, Curt A, Demkó L. Towards a Mobile Gait Analysis for Patients with a Spinal Cord Injury: A Robust Algorithm Validated for Slow Walking Speeds. Sensors Basel Switz 21, 7381 (2021)
Wu J, Kuruvithadam K, Schaer A, Stoneham R, Chatzipirpiridis G, Easthope CA, Barry G, Martin J, Pané S, Nelson BJ, Ergeneman O, Torun H. An Intelligent In-Shoe System for Gait Monitoring and Analysis with Optimized Sampling and Real-Time Visualization Capabilities. Sensors Basel Switz 21, 2869 (2021)
Wu J, Maurenbrecher H, Schaer A, Becsek B, Easthope CA, Chatzipirpiridis G, Ergeneman O, Pane S, Nelson BJ. Preprint Human Gait-labeling Uncertainty and a Hybrid Model for Gait Segmentation. (2021) doi:10.36227/techrxiv.16924102.v1
Bannwart M, Bayer SL, König Ignasiak N, Bolliger M, Rauter G, Easthope CA. Mediolateral damping of an overhead body weight support system assists stability during treadmill walking. J Neuroeng Rehabil 17, 108 (2020)
Swanenburg J, Langenfeld A, Easthope CA, Meier ML, Ullrich O, Schweinhardt P. Microgravity and Hypergravity Induced by Parabolic Flight Differently Affect Lumbar Spinal Stiffness. Front Physiol 11, 562557 (2020)
Stahl C., Gateau B. Ferrini K. Experiments on the localisation of cooking recipes content using semantic food descriptions. 2020 15th Int Work Semantic Soc Media Adapt Personalization Sma 00, 1–5 (2020).
Melendez-Calderon A, Shirota C, Balasubramanian S. Estimating Movement Smoothness from Inertial Measurement Units. Biorxiv 2020.04.30.069930 (2020)
Bannwart M, Rohland E, Easthope CA, Rauter G, Bolliger M. Robotic body weight support enables safe stair negotiation in compliance with basic locomotor principles. J Neuroeng Rehabil 16, 157 (2019)
Shirota, C., Balasubramanian, S. & Melendez-Calderon, A. 2019. Technology-aided assessments of sensorimotor function: current use, barriers and future directions in the view of different stakeholders. J Neuroeng Rehabil 16, 53 (2019)
Widmer M, Lutz K, Luft AR. 2019. Reduced striatal activation in response to rewarding motor performance feedback after stroke. Neuroimage Clin 24, 102036 (2019)
Giovanoli, S., Werge, T. M., Mortensen, P. B., Didriksen, M. & Meyer, U. Interactive effects between hemizygous 15q13.3 microdeletion and peripubertal stress on adult behavioral functions. Neuropsychopharmacol 44, 703–710 (2019)
Widmer M, Samara Stulz S., Luft AR, Lutz K. Elderly adults show higher ventral striatal activation in response to motor performance related rewards than young adults. Neurosci Lett 661, 18–22 (2017)
Radder B, Prange-Lasonder G, Kottink AIR, Melendez-Calderon A, Buurke JH, Rietman JS. Feasibility of a wearable soft-robotic glove to support impaired hand function in stroke patients. J Rehabil Med 50, 598–606 (2018)
Lutz K. Functional brain anatomy of exercise regulation. Prog Brain Res 240, 341–352 (2018)
New Publication
Have a look at our recent publication on wearable sensors to estimate movement quality in stroke by our Dr. Josef Schönhammer and Dr. Chris Awai.
New Publication
Have a look on our recent publication on Reward during Arm Training after Stroke by our Dr. Mario Widmer and Dr. Kai Lutz Widmer M,