Ambulatory assessment of motor performance after stroke

Hele tekst

(1)AMBULATORY ASSESSMENT OF MOTOR PERFORMANCE AFTER STROKE. AMBULATORY ASSESSMENT OF MOTOR PERFORMANCE AFTER STROKE. FOKKE VAN MEULEN. FOKKE VAN MEULEN.

(2) AMBULATORY ASSESSMENT OF MOTOR PERFORMANCE AFTER STROKE. Fokke van Meulen.

(3) Faculty of Electrical Engineering, Mathematics and Computer Science Biomedical Signals & Systems Institute for Biomedical Technology and Technical Medicine P.O. Box 217, 7500 AE, Enschede, the Netherlands. The research described in this thesis is part of the INTERACTION project, which is partially funded by the European Commission under the 7th Framework Programme (FP7-ICT-2011-7-287351) and coordinated by prof. dr. ir. P. H. Veltink, University of Twente, Enschede, the Netherlands. Financial support for printing of this dissertation was kindly provided by Xsens Technologies B.V.. Paranymphs:. Oebele van der Veen & Robert-Jan Doll. Cover:. ProefschriftOntwerp.nl, Bregje Jaspers. Printing:. Ridderprint B.V., Ridderkerk. ISBN:. 978-90-365-4381-1. DOI:. 10.3990/1.9789036543811. © F. B. van Meulen, 2017 – All rights reserved No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage or retrieval system, without written permission from the author..

(4) ambulatory assessment of motor performance after stroke. dissertation. to obtain the degree of doctor at the University of Twente, on the authority of the rector magnificus, prof. dr. T. T. M. Palstra, on account of the decision of the graduation committee, to be publicly defended on Friday the 22nd of September 2017, at 12:45. by. Fokke Bastiaan van Meulen. born on the 13th of February, 1987 in Rotterdam, the Netherlands.

(5) This dissertation has been approved by: Supervisor:. Prof. dr. ir. P. H. Veltink. Co-supervisor:. Dr. J. H. Buurke. © F. B. van Meulen, 2017 ISBN: 978-90-365-4381-1.

(6) Composition of the Graduation Committee: Chairman and secretary: Prof. dr. ir. P. M. G. Apers. University of Twente. Supervisor: Prof. dr. ir. P. H. Veltink. University of Twente. Co-supervisor: Dr. J. H. Buurke. University of Twente, Roessingh Research and Development. Members - internal: Prof. dr. ir. H. van der Kooij. University of Twente. Prof. dr. ir. H. J. Hermens. University of Twente, Roessingh Research and Development. Members - external: Prof. dr. J. P. Dewald. Northwestern University. Prof. dr. A. C. H. Geurts. Radboudumc. Dr. J. B. J. Bussmann. Erasmus MC.

(7)

(8) Contents. 1 Introduction. 1. 1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Assessment related to the ICF model . . . . . . . . . . . . . . . . 1.2.1 Objective evaluation of motor performance in a lab . 1.2.2 Wearable sensing . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4 Thesis goals and outline . . . . . . . . . . . . . . . . . . . . . . . . .. Part 1:. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. 3 4 5 7 10 12. Assessment of upper extremity function. 2 Assessment of daily life reaching performance 2.1 Introduction . . . . . . . . . . . . . . . . . . 2.2 Materials and Methods . . . . . . . . . . 2.2.1 Participants . . . . . . . . . . . . . 2.2.2 Clinical Assessment of Stroke . 2.2.3 Equipment . . . . . . . . . . . . . . 2.2.4 Equipment validation. . . . . . . 2.2.5 Protocol . . . . . . . . . . . . . . . . 2.2.6 Data Analysis . . . . . . . . . . . . 2.2.7 Statistical Analysis . . . . . . . . 2.3 Results . . . . . . . . . . . . . . . . . . . . . . 2.3.1 Equipment validation. . . . . . . 2.3.2 Simulated in-home task . . . . . 2.4 Discussion . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. 17 . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. 3 Assessment of arm movements using one inertial sensor 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.1 System setup. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2 Sensor orientation, velocity and position estimation. 3.2.3 Reduction of signal drift . . . . . . . . . . . . . . . . . . . . 3.2.4 Validation protocol . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 19 20 20 20 21 21 22 22 24 24 24 24 27 31. . . . . . . . .. . . . . . . . .. . . . . . . . .. . . . . . . . .. . . . . . . . .. . . . . . . . .. . . . . . . . .. 33 34 34 35 36 37 39 42.

(9) Part 2:. Assessment of lower extremity function. 4 Ambulatory assessment of walking balance 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Materials and Methods . . . . . . . . . . . . . . . . 4.2.1 System setup. . . . . . . . . . . . . . . . . . . 4.2.2 Participants . . . . . . . . . . . . . . . . . . . 4.2.3 Experimental protocol . . . . . . . . . . . . 4.2.4 Data processing. . . . . . . . . . . . . . . . . 4.2.5 Data analysis . . . . . . . . . . . . . . . . . . 4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 4.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . .. . . . . . . . . . .. 5 Analysis of balance during functional walking 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Materials and Methods . . . . . . . . . . . . . . . . . . 5.2.1 Ethical approval . . . . . . . . . . . . . . . . . . 5.2.2 Measurement Setup . . . . . . . . . . . . . . . . 5.2.3 Evaluation of dynamic balance . . . . . . . . 5.2.4 Experimental protocol . . . . . . . . . . . . . . 5.2.5 Data processing. . . . . . . . . . . . . . . . . . . 5.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.1 Metric evaluation . . . . . . . . . . . . . . . . . 5.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . .. . . . . . . . . . .. . . . . . . . . . .. . . . . . . . . . .. . . . . . . . . . .. . . . . . . . . . .. . . . . . . . . . .. . . . . . . . . . .. . . . . . . . . . .. . . . . . . . . . .. . . . . . . . . . .. . . . . . . . . . .. . . . . . . . . . .. 47 49 50 50 51 52 52 55 57 61 64. . . . . . . . . . . .. . . . . . . . . . . .. . . . . . . . . . . .. . . . . . . . . . . .. 65 67 68 68 68 69 73 73 74 79 82 85. in daily life .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... ........... . . . . . . . . . . . . .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. 91 93 94 94 96 98 99 101 102 103 103 103 103 107. . . . . . . . . . . .. . . . . . . . . . . .. . . . . . . . . . . .. . . . . . . . . . . .. . . . . . . . . . . .. . . . . . . . . . . .. . . . . . . . . . . .. . . . . . . . . . . .. . . . . . . . . . . .. Part 3: Evaluation of movement data 6 Objective evaluation of the quality of movement 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Materials and Methods . . . . . . . . . . . . . . . . . . . . 6.2.1 Requirement analysis . . . . . . . . . . . . . . . . . 6.2.2 Design of metrics. . . . . . . . . . . . . . . . . . . . 6.2.3 Sensor system overview . . . . . . . . . . . . . . . 6.2.4 Development of an activity monitor . . . . . . 6.2.5 Presenting large amounts of data . . . . . . . . 6.2.6 Patient evaluation . . . . . . . . . . . . . . . . . . . 6.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.1 Activity monitor . . . . . . . . . . . . . . . . . . . . 6.3.2 Lower extremity results . . . . . . . . . . . . . . . 6.3.3 Upper extremity results . . . . . . . . . . . . . . . 6.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..

(10) 7 General discussion 7.1 The INTERACTION system . . . . . . . . 7.2 Assessment of upper extremity function 7.3 Assessment of lower extremity function 7.4 Evaluation of movement data. . . . . . . . 7.5 General conclusion . . . . . . . . . . . . . . . 7.6 Future perspective . . . . . . . . . . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. 109 111 112 114 116 117 117. References. 121. Summary. 133. Samenvatting. 137. Dankwoord. 143. Biography. 147. List of publications. 149.

(11)

(12) CHAPTER 1. Introduction. 1.

(13) 2. 1. |. Chapter 1.

(14) Introduction. 1.1. |. 3. Background. Stroke, or cerebrovascular accident, is a disruption of blood supply to brain tissue which results in irreversible cell death. The two most common types of stroke are ischemic and hemorrhagic stroke. Ischemic stroke is a blood clot or other type of embolus in an artery reducing or completely blocking the artery and thereby preventing perfusion of distal brain tissue. Hemorrhagic stroke is a rupture of an artery in the brain that results in the accumulation of blood which may disrupt or compress adjacent brain tissue. Both types of stroke may cause many variations of disabilities depending on size and position of the brain which is affected. Stroke is the third most common cause of death in developed countries and is currently exceeded only by ischaemic heart diseases and cancer [32, 87]. Every year, around 50.000 people are affected by stroke in the Netherlands [100] and the worldwide incidence of stroke is over 15 million people [87]. Common risk factors of stroke are physical inactivity, high blood pressure, high blood cholesterol, tobacco use, atrial fibrillation, unhealthy diet and diabetes [87]. Overall, stroke is the leading cause of serious and long-term disabilities in developed countries [1, 32, 101]. Of those who survive a stroke, common disabilities are upper and lower extremity motor deficits, cognitive dysfunction, incontinence, dysphagia and dysphasia [75]. This thesis focuses on deficits in motor performance caused by a stroke. Depending on the patient’s remaining motor performance as a result of the stroke event, a patient-specific rehabilitation program is initiated. The first stage of the rehabilitation program includes intensive training sessions. During this phase, patient’s capacity to ambulate, ability to use his or her arms and other body functions are regularly evaluated. When the functional abilities of the patient are sufficient to live at home, the patient is discharged from the rehabilitation center and sent home. Just before discharge, a training program for the second stage of the rehabilitation program is made. This training may include outpatient treatments, occupational therapy, daily exercises and the advice to use the affected body part as much as possible. Training and advises are intended to preserve and improve the motor function of the affected body. Despite this rehabilitation program, patients regularly show deterioration of motor function on an activity level at the end of the second stage or after a certain amount of time [162]. This deterioration often results in an expensive rehospitalisation or institutionalization [2]. This might be avoided if the stroke survivor’s adherence to the training program and the motor performance during daily life would be better observable by a physician. Currently, they do not have any objective information available about. 1.

(15) |. 4. Chapter 1. the patient’s adherence to a prescribed training program and given advice. In order to explain any functional progress or decline of motor function, more information on body movements during daily life activities is necessary. By monitoring stroke survivors during their daily live activities, this objective information on motor function can be obtained. This information can be essential for an optimal. 1. guidance of rehabilitation therapy. However, todays systems merely evaluate the frequency and type of performed activities, but not how stroke survivors performed these movements [15]. This thesis focuses on the assessment of motor function, in particular of those who are affected by a stroke. Using a wearable sensing system, movements of upper and lower extremities can be monitored in a daily life setting. For this purpose, algorithms need to be developed for the estimation and evaluation of motor function.. 1.2. Assessment related to the ICF model. Assessment of human functioning is part of daily practice in rehabilitation care. Human functioning is frequently assessed using standardized clinical tests and/or questionnaires. Any found deficits in for instance extremity function, functional abilities or social life can be classified accordingly to the International Classification of Functioning, Disability and Health (ICF). The ICF is the framework of the World Health Organization that describes the functional state of an individual and is used in different medical sectors [163]. Within the ICF, a biopsychosocioal model is presented to describe a patient’s functioning on a level of body structure and body function, activity and participation (Fig. 1.1). The activity level is frequently subdivided in capacity, which describes what a person can do, and performance, what a person actually does. It is a holistic model in which all human functioning takes place in a certain context, taking into account environmental and personal factors. Standardized clinical assessments are frequently used to assess a person’s motor performance on the level of body function or on the level of activity. On the level of body function, these assessments are used to evaluate for instance, a person’s ability to passively or actively make a certain movement. Such tests, typically describe the maximum movement a person is able to perform, as part of the normal movement range as in healthy subjects. On the activity level, these tests are used to evaluate a person’s ability to complete specific activities of daily living or other functional tasks. The results of these assessments on activity level typically describe a person’s ability to complete a task, the time needed to complete a task or the.

(16) Introduction. |. 5. Health condition. Body structure Body function. Activity Performance / Capacity. Participation. 1 Environmental factors. Personal factors. Figure 1.1: Health model in the ICF framework. number of times a task can be completed within a certain time period. Although, these clinical assessment scales describe which body movements can be performed and which activities can be completed, they do not describe the quality of the body movements which are actually performed while completing an activity. Thereby it remains unknown which body functions are actually used during a certain activity. As a consequence, it remains difficult or maybe impossible to describe any progress or deterioration of motor performance during daily life activities, based on the results of standardized clinical tests. For example: if during a follow up test a patient is able to complete the activity in a shorter period of time, this may result in a higher test score. However, it is unclear how the patient reached this higher test score. Was the patient’s higher test score the result of restoration of body function? Was it the result of a higher performance within the ranges of a person’s capacity? or was the patient somehow compensating for a lack of body function? Therefore, more objective and qualitative evaluation methods of motor performance during daily life activities are needed for a clinical assessment of functional tasks on body function level. Results of these assessment methods should be patient-specific and directly applicable in rehabilitation practice. 1.2.1. Objective evaluation of motor performance in a lab. In the past decades, many systems were developed for the objective evaluation of motor performance. Nowadays, specialized movement laboratories are available for the objective assessment of all kind of human body movements. Most common systems in movement laboratories are used to describe the kinematics and kinetics of human body movements within a lab environment..

(17) 6. |. Chapter 1. Evaluation of kinematics in a lab environment Optical reference systems are assumed to be the ‘gold standard’ in the assessment of positions of the human body, i.e., kinematics. Many systems use an on-body marker(set) which actively sends a light signal or passively reflects light from an external source. Light from the markers is captured by multiple cameras around. 1. the person. By knowing the relative position of all cameras, the system can estimate the relative position of a marker(set). These systems can be highly accurate (errors < 1 cm), however, there are some disadvantages when using optical systems. For example, for an accurate estimation of the relative position of a marker, the marker should always be visible by at least three cameras. If there is a line of sight problem (i.e., there is something between the cameras and the marker) the marker position can not (accurately) be estimated. Which might be caused by a piece of clothing, a piece of furniture or another body part, that reduces the visibility of a marker. Furthermore, when the system is calibrated and ready to use, the measurement space is limited to the calibrated area between the cameras. Any movement outside the measurement space cannot be estimated. Any movement of a camera after calibration, will also cause inaccurate marker position estimations. Evaluation of kinetics in a lab environment Kinetics of the human body are interacting forces and torques between the body and its environment. These kinetic parameters describe biomechanical quantities of an individual. Such as, the center of pressure (CoP), i.e., the point of application of the ground reaction force, and the center of mass (CoM), i.e., point where all mass of an object is concentrated. These measures relative to a person’s base of support, are frequently related to body balance during standing and walking [54]. Force plates, sensorised treadmills or sensorised walkways are commonly used for the measurement of one or three-dimensional forces and torques. Again, these lab-based systems can be highly accurate, but there are some disadvantages while using these measurement systems. When using a force plate, a person often needs to adapt his or her way of walking to ‘hit’ the plate. An incomplete contact with the force plate will cause an incomplete registration of ground reaction forces and inaccurate estimation of CoP and CoM positions. Although sensorised treadmills and walkways show a complete registration of ground reaction forces, only straight walking patterns can be evaluated. Therefore, kinetic evaluation in a lab setting can barely represent real daily life situations..

(18) Introduction. |. 7. General disadvantages of lab-based systems Body movements can accurately be monitored within movement laboratories, however, those laboratories are not generally available for daily life practice. Used equipment is expensive, unmovable and can only be used by specialized personnel. Furthermore, a person’s performance can only be limited evaluated by simulating a daily life setting. To overcome these disadvantages of lab-based systems, there is an increasing effort in developing and applying wearable sensing systems. These wearable sensing systems generally are cheaper, in potential easier to use, and allow for ambulatory assessment of human body movements. 1.2.2. Wearable sensing. Ambulatory assessment of human body movements using wearable sensors, is the assessment of people’s body movements in a natural or unconstrained setting during activities of daily living. Evaluation and monitoring of body movements in a daily life setting is especially of interest in those with a limited body function due to severe brain injury (e.g., stroke) or chronic disease (e.g., Parkinson’s disease). Ambulatory assessment potentially allows the evaluation of effectiveness of any applied therapy in a real daily life practice. Wearable sensing systems for the assessment of human body movements are available in various shapes and sizes, depending on the objective of the measurement and the parameters of interest. Inertial sensing for evaluation of kinematics Wearable sensor systems for the evaluation of human body movements frequently contain three types of sensors: accelerometers, to measure sensor and gravitational acceleration; gyroscopes, to measure sensor angular velocity; and magnetometers, to measure direction of magnetic north. Especially accelerometers are frequently used for the assessment of human body movements. These sensors are small, relatively cheap and easy to integrate. Currently, almost every smart electronic device has an integrated accelerometer. Accelerometers are used for varying applications such as the estimation of the orientation of a body part, body gesture, or to count the number of steps made within a certain time interval. Signals of accelerometers and gyroscopes can be combined in one sensing system, a so called inertial sensor. Using both signals and specialized algorithms, sensor orientation, change of orientation as well as change of position can be estimated. Inertial sensors combined with magnetometers even allow the estimation of orientation relative to the magnetic north.. 1.

(19) 8. |. Chapter 1. Examples of wearable sensor systems that contain or combine accelerometers, gyroscopes and/or magnetometers are: • Pedometers. Pedometers are accelerometers integrated in a watch, as a clip attached to trousers, or integrated in a smartphone. Pedometers quantify. 1. human body movement by evaluating data of an accelerometer. It estimates the number of steps made within a certain time frame. Combined with information of the user, it is able to estimate walked distance as well as energy expenditure [24]. • Activity monitor. In most cases, one inertial sensor is used to evaluate the quantity of activity: how frequent is the user moving a certain body part, or which activity is the user performing. Depending on the body part the sensor is attached to, results can for instance be the number of reaches within a certain time frame, body gestures over time, percentage of time somebody is sitting, standing or walking [14, 76, 114, 151] • Motion capture. A motion capture system based on inertial sensors uses multiple sensors to evaluate movements of more than one body part at the same moment. By using two inertial sensors attached to two body parts which are connected to each other, the angle between both body parts can be estimated. For instance, by using a sensor on the upper and lower leg, a knee angle can be estimated [38]. When using a sensor on multiple body parts in combination with a bio-mechanical model, it is possible to estimate full three-dimensional body positions over time [124]. This motion capture technique is frequently used for animations in movies and computer games, but nowadays also used in sport sciences and rehabilitation medicine. Force sensing for evaluation of kinetics While standing on the ground, sitting in a chair, leaning towards something, or any other activity, the human body interacts with the environment. By using one or more force sensors, the interacting forces between the human body and the environment can be measured. Wearable force sensors have been developed for the evaluation of standing and walking and measure or estimate the ground reaction force (GRF) between the human body and the ground. These sensors are equipped inside the shoe, underneath the shoe or within a walking aid. Two examples of wearable sensor systems for evaluation of kinetics are:.

(20) Introduction. |. 9. • Insoles. An insole is a shoe insert, containing multiple pressure sensors to evaluate pressure patterns during standing and walking. Accuracy depends on the quality and the number of sensors integrated in the insole. Using measured pressure patterns, CoP movement per foot can be estimated. The main advantage of insoles is the option to integrate this measurement system inside a person’s own pair of shoes. It allows the evaluation of multiple steps without any restriction on foot placement. However, the integrated pressure sensors only evaluate forces in one direction, the vertical component of the GRF. Advanced algorithms or additional sensing is necessary for an estimation of all components of the GRFs [23, 78]. Another disadvantage of many insoles is the inability to use the sensors for a longer period of time. Furthermore, the insole should perfectly fit inside the shoe, otherwise measured pressure cannot be accurately estimated. • Instrumented force shoes. Instrumented force shoes are sandales equipped with sensors, inside or underneath the shoe, such as Xsens’ instrumented ForceShoe™ (Fig. 1.2). Forces between the shoes and the ground can be measured using two three-dimensional force and moment sensors. Advantage of this method is the possibility to accurately estimate the three-dimensional ground reaction forces and estimate the CoP and CoM, for many steps and during many types of walking patterns, as long as there is no other contact with the environment [133]. The main disadvantage is the increased sole height and weight of the shoe compared with normal shoes. Although the influence on the walking pattern is limited [106], walking with the shoes is more exhausting compared with regular shoes.. Figure 1.2: Xsens’ instrumented ForceShoe™. A sandal with a three-dimensional force and moment sensor as well as an inertial sensor underneath the heel and forefoot segment.. 1.

(21) |. 10. 1.3. Chapter 1. Challenges. Most wearable sensing systems for the assessment of motor performance are used to quantify instead of qualify body movements. These sensing systems report metrics such as number of steps, number of times a person is reaching, number of sit-. 1. to-stand transitions, number of falls. Metrics which are of interest, but do not evaluate impairment levels during functional tasks. Furthermore, these metrics do not describe on the level of body function how a person is walking, reaching, raising up from a chair or why a person is falling while he or she is performing different activities of daily living. Figure 1.3 relates previously described systems for the evaluation of motor performance in terms of laboratory or clinical settings versus daily life settings and evaluation of quantity versus quality of body movements. Currently, as presented in this overview, there is almost no sensor system available for the qualitative assessment of human body movements in a daily life setting (upper right corner of the figure). More research is needed to develop a wearable sensor system for the unobtrusive and qualitative assessment of daily life movements of stroke survivors. Quality of movement Force plate Optical reference system Instrumented shoes Fugl-Meyer assessment Multiple inertial sensor system. Laboratory or clinical setting. Berg Balance Scale. Daily life setting. Activity monitor. Timed Up and Go test 10 meter walk test. Fall detector Pedometer. Quantity of movement. Figure 1.3: Different methods and systems for the evaluation of motor performance in terms of laboratory versus daily life setting (left - right) and evaluation of quality versus quantity of body movements (top - bottom). The goal of this thesis and the European project INTERACTION, was to primary reach the upper right corner, a wearable sensing system for the qualitative evaluation of body movements..

(22) Introduction. |. 11. INTERACTION project In November 2011, the research project ‘INTERACTION’ started the development of a wearable sensing system for the unobtrusive and qualitative assessment of daily life movements. This project was partly funded by the European Union under the 7th Framework Programme and was a collaboration of several international partners. From the Netherlands: the Biomedical Signals and Systems group of the University of Twente, Roessingh Research and Development and Xsens Technologies B.V.; from Switzerland: University of Zurich, and from Italy: Smartex S.r.l. and University of Pisa. The aim of the INTERACTION project was to develop and validate an unobtrusive and modular system for monitoring the quality of upper and lower extremity motor function in stroke survivors during daily life activities. This project aim was subdivided into several project objectives: 1) specifying the INTERACTION system, 2) development of a textile integrated sensing system, 3) development of algorithms to evaluate movement data, 4) development of methods to present large amounts of movement data, and 5) evaluation of developed sensing system and evaluation methods in clinical and daily life setting. Questionnaires, interviews and consensus meetings with stroke survivors and physicians were performed, to evaluate the user requirements and expectations about a monitoring system for rehabilitation. Based on these requirements a true ambulatory sensor suit integrated in textile and patient shoes was developed in the INTERACTION project (Fig. 1.4).. Figure 1.4: INTERACTION suit - Modular and wearable sensor suit for the qualitative assessment of motor performance of stroke survivors [66].. 1.

(23) |. 12. Chapter 1. The proposed sensor suit is part of the INTERACTION system, as presented in Fig. 1.5. The sensor suit contains multiple sensors and generates large amounts of data. Although the health care professional is interested in all registered movements performed during daily life, this professional is not able to review all data. The remaining question is therefore: how to define, estimate and present clinically. 1. relevant metrics for the evaluation of upper and lower extremity motor function of stroke survivors who are performing activities in a daily life setting? Such that it reduces the amount of data, which allows the health care professional to quickly interpret the clinically useful metrics. Measuring. Analysis. Feedback. Evaluation Figure 1.5: Model of the INTERACTION system - Using the INTERACTION wearable sensing system, movement data are measured and send wirelessly to a home portal. Via a secure internet connection, data are send from the home portal to a database. Algorithms assess all data and generate reports of each measurement session. Health care professionals evaluate the reports and if necessary review the data and generate new reports. Finally, the health care professional can give feedback to the patient, directly or via a home portal. Dashed arrows are pathways which are not realized in the INTERACTION project.. 1.4. Thesis goals and outline. The main goal of this thesis is to develop clinically relevant metrics for the assessment of stroke survivors’ quality of upper and lower extremity motor performance and estimate them from on-body sensing information measured during daily life using the INTERACTION sensing system..

(24) Introduction. |. 13. Based on the modular design of the INTERACTION sensor suit and the model of the INTERACTION system, the main goal is subdivided into three parts: • Part I: is focusing on the development of algorithms for the assessment of upper extremities movements of stroke survivors who are performing multiple arm tasks in a daily life setting. In Chapter 2, a method for the assessment of arm movements in a daily life setting is proposed. Metrics such as reaching area and reaching distance were related to a frequently used clinical assessment scale. Chapter 3 is focusing on the assessment of arm movements with only one inertial sensor. It describes which metrics can be assessed under which conditions, with a minimal sensor set. • Part II: is focusing on the development of algorithms for the assessment of movements of the lower extremities and walking balance of stroke survivors who are ambulating in a daily life setting. In Chapter 4, data processing methods for the evaluation of walking balance while walking in straight line are proposed. The first prototype of the INTERACTION system was used to evaluate kinematics and kinetics of every step. In the second chapter of this part, Chapter 5, walking balance is evaluated while walking straight as well as when turning. • Part III: is focusing on the development of data processing methods for the processing of larger amounts of human body movement data into metrics for an orderly assessment. This is represented in Fig. 1.5, as the process from measuring to evaluation. Chapter 6, describes the iterative process of metric development towards one page reports of the upper and lower extremities for the evaluation of the quality of human body movements. These reports allow the assessment of intra-patient differences of their quality of movements while performing different types of tasks. This thesis finishes with a general discussion of all results presented in Chapter 2 to 6. Finally, some suggestions and an outlook for future research are given.. 1.

(25)

(26) PART I. ASSESSMENT OF UPPER EXTREMITY FUNCTION.

(27)

(28) CHAPTER 2. 2 Assessment of daily life reaching performance after stroke. Published by Springer US as: F. B. van Meulen, J. Reenalda, J. H. Buurke, P. H. Veltink (2015). Assessment of dailylife reaching performance after stroke. Annals of biomedical engineering, 43(2), 478–486. DOI: 10.1007/s10439-014-1198-y.

(29) 18. |. Chapter 2. Abstract. 2. For an optimal guidance of the rehabilitation therapy of stroke patients in an inhome setting, objective, and patient-specific performance assessment of arm movements is needed. In this study, metrics of hand movement relative to the pelvis and the sternum were estimated in 13 stroke participants using a full body ambulatory movement analysis system, including 17 inertial sensors integrated in a body-worn suit. Results were compared with the level of arm impairment evaluated with the upper extremity part of the Fugl-Meyer Assessment scale (uFMA). Metrics of arm movement performance of the affected side, including size of work area, maximum reaching distance and movement range in vertical direction, were evaluated during a simulated daily life task. These metrics appeared to strongly correlate with uFMA scores. Using this body-worn sensor system, metrics of the performance of arm movements can easily be measured and evaluated while the participant is ambulating in a simulated daily life setting. Suggested metrics can be used to objectively assess the performance of the arm movements over a longer period in a daily life setting. Further development of the body-worn sensing system is needed before it can be unobtrusively used in a daily life setting. Key words: Ambulatory Assessment, Arm Tasks, Fugl-Meyer List of abbreviations: uFMA. Upper extremity part of the Fugl-Meyer assessment scale (0-66 points). IMU. Inertial measurement unit.

(30) Assessment of daily life reaching performance. 2.1. |. 19. Introduction. Performance of several activities of daily living depends on proper arm function. Many stroke patients have a reduced ability to coordinate their arm movements. Intensive rehabilitation therapy is usually given to enhance recovery and studies have shown that this can increase arm motor recovery [36, 144]. For an optimal guidance of the rehabilitation therapy, medical professionals need frequently measured and patient-specific information of arm function [68]. Currently, arm function of stroke patients is assessed according to the international classification of functioning [46]. In a clinical setting, this assessment is usually done on three levels. First, the assessment is performed on the level of body function, i.e., level of arm impairment (e.g., Fugl-Meyer Assessment [42]). Secondly and thirdly, the assessment is done on the level of activities and participation, i.e., assessment of prescribed arm tasks (e.g., action reached arm test [86], stroke upper limb capacity scale [125]). However, it remains largely unknown whether the clinically assessed level of arm impairment, using for instance the uFMA, reflect the actual performance of arm movements in daily life. Information of arm use in a daily life setting will allow assessment of the transfer of learned arm movements to daily life performance [120]. Previous studies report methods for quantitative and qualitative assessment of upper arm movements in stroke subjects [29, 30, 35, 36, 63, 71, 93, 99, 143]. For instance, Kamper et al. report a decreasing active range of motion with increasing severity of impairment, in seated stroke subjects under the condition of restricted trunk motion [63]. Subramanian et al. suggest that movement quality metrics of trunk displacement and shoulder flexion are more sensitive in identifying upper extremity deficits than clinical scales [143]. Many of these kinematic studies have been performed in laboratory settings, with specialized laboratory-bound equipment [30, 35, 36, 63, 71, 93, 99, 143]. However, for assessment of movement performance of the arm in a daily life setting, a wearable system with a minimized impact on normal behavior is preferred [9]. A feasible method for movement assessment in a daily life setting is the use of inertial measurement units (IMUs). IMUs are small, unobtrusive and can easily be worn on the body. The main advantage of IMUs is that an external physical reference system is not required. This allows easy assessment of movements in a daily life setting [14, 170]. IMUs are already used for the evaluation of the upper extremity movements in terms of joint angles and end-point position, speed, acceleration and smoothness of the hands [84, 112, 120, 147, 149]. When using multiple inertial sensors, the orientation and movement of different parts of the body can be evaluated. Furthermore, with the addition of precisely measured body dimensions and a biomechanical model of the human body segments, relative positions of all instrumented extremities can be estimated using linked kinematic chain models [124]. Long-term daily life measurements may result in large amounts of kinematic data. Well-chosen metrics are needed to allow objective evaluation of arm movement perfor-. 2.

(31) |. 20. Chapter 2. mance in a daily life setting and provide objective information about motor strategies associated with specific tasks [120]. These metrics may include maximum reaching distance, traveled path length and work area in multiple planes. Furthermore, hand position can be evaluated relative to the sternum as well as to the pelvis, to study compensatory trunk movement strategies. As a consequence of different (compensatory) strategies [18], hand-sternum and hand-pelvis distances may vary differently between subjects. The objective of this study is to evaluate metrics that describe daily life arm movement performance in stroke subjects of both the affected and the unaffected arm in a simulated daily life setting. These metrics were derived using a body-worn inertial sensing system. 2. and related to participant’s scores of the frequently used upper extremity part of the FuglMeyer assessment scale in order to assess whether daily life arm movement performance relates to clinically-assessed level of impairment.. 2.2 2.2.1. Materials and Methods Participants. Seventeen stroke participants were recruited from Roessingh rehabilitation centre, located in Enschede, the Netherlands. Recruited participants were between 35 and 75 years of age and had a hemiparesis as a result of a single unilateral stroke, diagnosed at least six months earlier. Furthermore, participants had to be able to lift their affected arm against gravity from a relaxed vertical orientation onto a table directly in front of them while seated. Exclusion criteria were a medical history with more than one stroke, inability to understand questionnaires and inability to perform given instructions. The study protocol is a subset of a larger protocol approved by the local medical ethics committee. Each participant signed a written informed consent before participating. Three participants with severely affected lower extremity function were not able to complete the task without assistance due to unstable walking patterns. The corresponding test results were excluded from the analysis. Data of one other participant was not fully recorded because of a broken cable during the session. Remaining were 13 participants with an average age of 63.9 (SD ± 9.0) years, 2.3 (SD ± 1.8) years post stroke and of which eight are male. Participant specific information is reported in Table 2.1 and includes age, number of years post stroke, height, weight and dominant side.. 2.2.2. Clinical Assessment of Stroke. Level of impairment of the affected side of each participant was assessed using the upper extremity part of the Fugl-Meyer Assessment Scale (uFMA). This scale ranges from zero to 66 points [42]. All assessments were performed by the same researcher with a background in technical medicine and adequate clinical expertise to perform the assessment..

(32) Assessment of daily life reaching performance. |. 21. Table 2.1: General participant characteristics ID1. Gender. s1 s2 s3 s4 s5 s6 s7 s8 s9 s10 s11 s12 s13. M M M M F M M M F F F F M. Age. 2. Post stroke2. Dominant side. Affected side. uFMA3. 7.4 4.0 1.8 2.4 3.3 1.3 1.6 1.6 0.7 1.4 1.4 1.6 1.2. R R R R R R R R R R R R L. L L L R L L L L L R L L L. 49 43 20 54 56 53 49 36 43 63 23 59 54. 70 69 47 73 67 65 75 52 60 71 55 56 70. 1 Participant. identification number. ment scale (0-66 points) [42].. 2.2.3. 2 In. years.. 3 Upper. extremity part of the Fugl-Meyer assess-. Equipment. Kinematic data were recorded with the MVN Biomech motion capturing system (Xsens Technologies B.V., Enschede, the Netherlands) at a frequency of 120 Hz [124]. To measure full body kinematics this system includes 17 IMUs positioned symmetrically on both sides, specifically at the feet, lower legs, upper legs, hands, lower arms, upper arms and shoulders. Additional sensors were positioned on sternum, sacrum and the head. To reduce movement artifacts sensors were mounted in plastic brackets strapped to the body and fixated to the skin using Velcro® . Sternum and shoulder sensors were attached using a small unobtrusive harness. Straps were tightened and wiring was tucked away behind clips, in the least obtrusive way. To check for any interfering straps or wires, participants were asked to simply walk around and strapping and wiring were adjusted whenever necessary. Data of all sensors, 3D acceleration, 3D rotation velocity and 3D earth-magnetic field direction and Xsens’ software MVN Studio Pro (version 4.2, Xsens Technologies B.V., Enschede, the Netherlands) were used to estimate body segment orientation, relative segment position and full body 3D kinematics [124]. A reference video was recorded to verify estimated 3D kinematics.. 2.2.4. Equipment validation. Inherent to using IMUs is the presence of positional drift caused by integration of acceleration and angular velocity signals. This occurs after a few seconds of measuring [47]. More accurate estimates of relative positions of body segments can be obtained using prior knowledge of segment lengths and assuming a linked kinematic chain based on known joint constraints [124].. 2.

(33) |. 22. Chapter 2. To address this, the accuracy of the position estimates of the MVN Biomech system was first validated against an optical reference system (Vicon, Oxford Metrics, Oxford, United Kingdom). This experiment was performed on one healthy participant wearing both the IMUs and reflective markers positioned at the wrists, sternum and sacrum. While seated in front of a wooden table, performed several circular arm movements at three heights above the table. At each height 15 circular motions were performed with the left hand and 15 with the right hand. Movements were done in sequence and pairs of left and right arm motions were combined in single measurements. For all measurements the Euclidean distances from the base of the hand to the sternum and the base of the. 2. hand to the pelvis, were estimated using the MVN Biomech system and compared with the distances measured with the optical reference system. Mean and standard deviation of the differences between both measurements systems were evaluated.. 2.2.5. Protocol. As a subset of a larger protocol, participants were asked to perform a simulated in-home task. Arm movements were assessed while the participants performed multiple daily life activities. The task was repeated three times. The task started with the participants seated on a wooden chair in front of a wooden table, with the chair completely pulled up to the table. Participants were instructed to stand up from the chair, walk around the table, open, walk through and close a hinged door and walk to another table in the second room. On this table, two identical small solid tubes were placed upright, both having a diameter of 4.5 cm, a height of 12 cm and a weight of 500 g. Participants were asked to grasp the first tube, lift and displace it to a marker on the other side of the table, a distance of 50 cm. Next, participants were asked to grasp the second tube from the same table and return it to the starting table in the first room, passing through the same door. The height of both tables was 75 cm. The complete task was explained and demonstrated prior to the experiment and step-by-step instructions were also given while the participant was performing the task. A schematic overview of the task is presented in Fig. 2.1. No explicit instructions concerning arm use were given to the participant while performing the simulated in-home task. Participants were free to use their affected as well as non-affected arm in any way they preferred. The only objective given to the participants was to complete the task. There was no time restriction imposed to fulfilling the task.. 2.2.6. Data Analysis. The performance of reaching movements was derived from quantitative analysis of arm and trunk movements. Kinematic data were processed offline by Xsens’ MVN studio Pro and yielded relative segment positions and orientations. Subsequently, data was exported to and analyzed using MATLAB® (version 2013b, MathWorks Inc., Natick, MA, USA)..

(34) Assessment of daily life reaching performance. |. 23. 2 1. Table 2. Table 1. Figure 2.1: Schematic top-down overview of the simulated in-home task. Participants start and finish at the first table (Table 1), walk along a hinged door, move the first tube (1) along the second table (Table 2) and take a tube (2) back to the first table.. Kinematic data were expressed in a pelvis coordinate frame ψ p as well in a sternum coordinate frame ψ s . Of both coordinate frames the positive y direction was defined in posterior-anterior direction, positive z in vertical upwards direction, and x-axis perpendicular to the y- and z-axis in a right-handed fashion. Likewise, positions of the most proximal side of both hands, Php and Phs , were expressed in the coordinate frames of the pelvis and the sternum. For any hand this is: 0. Php = Rgp × (Phg − Ppg ) 0. Phs = Rgs × (Phg − Psg ) Where frame,. Ppg. Phg. (2.1) (2.2). is the position of the proximal side of the hand in the global coordinate. and Psg are the positions of the pelvis and the sternum in the global coordinate 0. 0. frame, and Rgp and Rgs are the transposed rotation matrices expressing the pelvis and the sternum in the global coordinate frame. Four metrics were used for evaluation of the performance of arm movements during the complete task. The first metric is the work area of each hand, estimated from the hand positions during the simulated in-home task and expressed in both the pelvis and sternum coordinate frames. More specifically, a Delaunay triangulation method was used, to create a two dimensional envelope around all positions of a hand in both the transversal and the sagittal plane. The areas of the envelopes were used as the estimated work areas. Secondly, to quantitatively assess hand movements in three-dimensions, the lengths of the 3D trajectories of both hands were evaluated. These trajectory lengths were estimated using the summation of Euclidean distances between consecutive positions of each hand expressed in the pelvis and the sternum coordinate frames. Thirdly, to qualitatively assess performance of arm reaching, largest reached distances between hand and pelvis as well as between hand and sternum were determined after projection of the hand movements in the transversal plane. Finally, the range of height differences between each hand and the pelvis and each hand and the sternum were estimated.. 2.

(35) |. 24 2.2.7. Chapter 2 Statistical Analysis. Results were averaged per participant and corrected for body height. Linear regression analysis was performed to estimate correlation of determination values between described metrics for both the affected and non-affected arms and uFMA scores.. 2.3 2.3.1. 2. Results Equipment validation. The 45 validation measurements were successfully recorded using both the MVN Biomech system and the optical reference system. The differences in distances estimated using both systems vary across all measurements and increase with hand-sternum distances. Averaged mean absolute differences of estimated distances evaluated over all measurements is 14 mm (SD ± 13 mm) for hand-sternum and 35 mm (SD ± 34 mm) for hand-pelvis. Over all measurements the mean of the largest absolute differences between both systems is 58 mm (SD ± 20 mm) for hand-sternum distance and 141 mm (SD ± 32 mm) for hand-pelvis distance.. 2.3.2. Simulated in-home task. All kinematic data were recorded without loss of data. Figure 2.2 shows typical hand position data with respect to the pelvis for a single participant (participant 3, uFMA score of 20 out of 66) performing the simulated in-home task. The thin traces show the trajectories projected on the transversal plane (xy-plane) as well as on the sagittal plane (yz-plane) of both hands relative to the position of the pelvis. The thicker green envelopes around the trajectories represent the overall largest reached distances in all directions in the transversal and the sagittal planes, while performing the simulated in-home task. Table 2.2 specifies the correlation of determination values for all four metrics evaluating hand positions relative to the pelvis and to the sternum for both the affected and unaffected side. All corresponding correlation coefficients were found to be positive, except for the largest reaching distance of the non-affected side, relative to the sternum in the transversal plane. Significant correlations with the uFMA scores were only found in the metrics evaluating hand movements of the affected side relative to the pelvis as well as relative to the sternum. The first correlating metric is the work area of the affected arm for movements in the transversal plane relative to the pelvis and the sternum (resp. r = 0.84, p < 0.001 and r = 0.70, p < 0.01), as well as in the sagittal plane relative to the pelvis and the sternum (resp. r = 0.84, p < 0.001 and r = 0.79, p < 0.01). Second correlating metric is the maximum reached distance in transversal plane relative to the pelvis and the sternum (resp. r = 0.88, p < 0.001 and r = 0.82, p < 0.001). The third correlating metric is the range in vertical hand elevation relative to the pelvis and the.

(36) Assessment of daily life reaching performance. |. 25. sternum (resp. r = 0.69, p < 0.05 and r = 0.76, p < 0.01). No significant correlations were found for the metric evaluating path length of the hand of the affected side relative to the pelvis as well as relative to the sternum. Figure 2.3 shows the results per patient of the three highest correlating metrics relative to their uFMA scores.. Right hand envelope area(x,y) = 0.377 m2 0.8. 0.6. 0.6. 0.4. 0.4. y (m). y (m). Left hand envelope area(x,y) = 0.0409 m2 0.8. 0.2. 0.2. 0. 0. −0.2. −0.2. −0.4 −0.6. −0.4. −0.2. 0. 0.2. 0.4. −0.4 −0.6. 0.6. 2. −0.4. −0.2. x (m) Left hand envelope area(y,z) = 0.0275 m2. 0.6. 0.6. 0.4. 0.4. 0.2. 0. −0.2. −0.2. 0.2. y (m). 0.6. 0.4. 0.6. 0.8. Envelope Positions. 0.2. 0. 0. 0.4. Right hand envelope area(y,z) = 0.141 m2. Envelope Positions. −0.2. 0.2. 0.8. z (m). z (m). 0.8. −0.4 −0.4. 0. x (m). −0.4 −0.4. −0.2. 0. 0.2. 0.4. 0.6. 0.8. y (m). Figure 2.2: Upper graphs – top-down view: transverse plane (xy-plane). Lower graphs – side view: sagittal plane (yz-plane). All showing arm positions relative to the pelvis (origin of the graph) during the simulated in-home task and corresponding envelope. Position data of a 28 s measurement. Participant number 3, uFMA = 20 out of 66..

(37) 26. |. Chapter 2. Table 2.2: Correlation of determination values (R2 ) between metrics and uFMA scores Relative to pelvis. Relative to sternum. Plane. A. NA. A. NA. xy. 0.70***. 0.03. 0.49**. 0.06. yz. 0.70***. 0.08. 0.62**. 0.17. Length of hand trajectories. xyz. 0.23. 0.01. 0.20. 0.05. Maximum reaching. xy. 0.77***. 0.07. 0.68***. 0.14†. 0.47*. 0.21. 0.58**. 0.25. Work area. 2. Range of hand elevation. z. xy-plane = transversal plane. yz-plane = sagittal plane. xyz-plane = three-dimensional. A = Affected side. NA = Non-affected side. * p < 0.05, ** p < 0.01, *** p < 0.001. † Corresponding correlation coefficient is negative. Results and corresponding linear models of underlined values are shown in Fig. 2.3.. 0.25 13. 0.15 9. 0.16 10 12. 7. 0.1 3. 0.12 13. 0.08. 0.04. 3. 8 11 0. 4 5 6. 7 1. 9. 2 1. 0.05. Area YZ (m2). 4 65. R2=0.77, p<0.001. 0.2. 10. 0.2. Area XY (m2). R2=0.7, p<0.001. 11. 8. 20. 40. 12. 2. 0 0. 20. 40. 60. 0. 60. Maximal reached distance XY (m). R2=0.7, p<0.001. 5 10 4 6 13 9 7 12 2 1. 0.6. 0.4. 3 8 11. 0.2. Fit Confidence bounds. 0 0. 20. 40. 60. Fugl−Meyer (0−66). Figure 2.3: Relation between uFMA scores and three metrics with the highest correlation of determination values for the simulated in-home task. First, mean area of the envelopes around movements of the affected arm in the transversal plane (xy-plane). Second, movements in the sagittal plane (yz-plane). Third figure is the maximal reached distance between pelvis and hand in the transversal plane (xy-plane)..

(38) Assessment of daily life reaching performance. 2.4. |. 27. Discussion. The objective of this study was to evaluate metrics which describe daily life arm movement performance in stroke subjects, using a body-worn inertial motion capture system. These metrics, estimated for a simulated in-home task, appeared to be related to participant’s uFMA scores, which express level of arm impairment. Significant correlations were found between uFMA scores and movement performance metrics estimated for the affected side in both the coordinate frames of the pelvis as well as the sternum. The uFMA scores appeared to significantly correlate with 1) the work area of the affected arm in the transversal and sagittal planes, 2) maximum reaching distance and 3) the range of vertical hand elevation. High correlation of determination values between these metrics and the uFMA scores show that the variance of the metrics is highly predictable from the uFMA scores. Corresponding correlations coefficients show strong positive relationships between these metrics and uFMA scores. These relationships show that stroke subjects with a higher uFMA score, that is a lower level of arm impairment, move their affected arm during a simulated in-home task over a larger area, a larger distance and a larger range of hand elevations than subjects with lower uFMA scores, representing a higher level of arm impairment. Since spasticity is a component of the uFMA this could at least in part account for the correlation in the data [18, 27, 35, 99]. Similar correlations were reported in stroke subjects performing more prescribed tasks, between uFMA scores and metrics describing kinematics of stroke subjects [30, 35, 36, 63, 71, 99, 143]. Strikingly, no relationships were found between the uFMA scores and any of the arm movement metrics evaluated for the non-affected arm. Such relationships could have been expected when assuming that severely affected subjects would compensate for the reduced movements of their affected side by enlarging the work area at their non-affected side. The absence of such relationships might be explained twofold. First, arm preference might have caused even less affected participants to move their non-affected arm more. The preferred arm is the non-affected arm in 10 out of the 13 participants that participated in the current study (Table 2.1). Secondly, a participant with a higher level of arm impairment could have an increased rotation of the pelvis in the transversal plane while completing the task, such that the required actions can be performed in the working area of the non-affected arm. This compensation strategy would not result in an increased reaching area or distance of the non-affected arm. Hand positions were evaluated relative to the sternum as well as relative to the pelvis. Differences between hand-sternum distances and hand-pelvis distances could have been expected as a result of compensatory trunk movements. Stroke subjects with a higher level of arm impairment will show less shoulder flexion and may compensate during reaching by trunk flexion [143]. Such compensatory movements of the trunk would increase handpelvis but not hand-sternum distance. This would result in higher correlations with uFMA scores for metrics describing movements of the hand relative to the sternum than relative. 2.

(39) 28. |. Chapter 2. to the pelvis. However, this was only found for the metric describing the range of vertical hand elevation. Therefore, our results do not indicate compensatory trunk movements during reaching in this simulated daily life task. To evaluate participant’s arm movements in a simulated daily life setting, no explicit instructions concerning arm use were given to the participant. As a consequence, participants could use different strategies to complete the prescribed task. For instance, participants walked through the same door in two directions and even though the hinge is on different sides while opening and closing the door, more affected participants opened and closed the hinged door in both directions with their non-affected side while several of. 2. the less affected participants used both arms. Another example of different strategies is the way participants moved a tube from one room to the other. Some of the participants kept the tube in the hand of their non-affected side at all time, while others moved the tube from their non-affected to their affected hand before opening the door with their non-affected side. Furthermore, a different impairment level of the hand can influence strategies. Using the MVN Biomech system, actual grasping cannot be detected and the way of grasping cannot be evaluated. Therefore, no distinction can be made between participants who are unable to and those who do not choose to use their affected hand for grasping. However, participants who have a lower uFMA score, especially for the hand evaluation parts, can be expected to apply alternative reaching and grasping strategies, avoiding using their affected arm and hand [18]. Positive correlations of the affected arm metrics with uFMA that were found may, therefore, be related to differences in applied reaching and grasping strategies. It should be noted that the considered simulated in-home task is of limited difficulty and may have been completed single-handed. The difficulty of the selected task may, therefore, have been of influence on the results of the metrics. Bimanual tasks, tasks with different object sizes or evaluating arm movements over a longer period may show different results. Validation measurements, performed in one healthy participant, show larger differences at larger reaching distances between the MVN Biomech system and the optical reference system. Differences are not random, but predictably related to reaching distance. These differences may be caused by incomplete registration of shoulder protraction and retraction and trunk movements by the MVN Biomech system. The shoulders and the back are not rigid body segments, therefore movements in these segments cannot completely be registered with the limited number of IMUs applied. Accuracy could be increased by using for instance additional sensing with goniometers on the spine and shoulder [81] or alternatively by fusing magnetic and inertial sensing [123]. Despite the deviations in distance estimation, assessed arm reaching appeared to be significantly correlated with uFMA in stroke subjects. Several limitations in the present work should be acknowledged. First, the task is performed in a simulated daily life setting; this setting will be different from the participant’s daily life setting. The participant might be unused to the setting and may apply different.

(40) Assessment of daily life reaching performance. |. 29. or non-optimal movement strategies compared with a real in-home setting. Secondly, no healthy control data of the simulated in-home task has been measured. Therefore, it remains unknown how much movement normally exists while completing the simulated in-home task. The best available control data in our study are the movements of the less affected participants performing the simulated in-home task. Our study demonstrates that these participants use their affected arm more extensively than the more impaired participants. Thirdly, motor performance varies over time within a single individual, as well as across different individuals performing the same task in different ways. We calculated the mean of three trials for each metric. This limited number of trials may have influenced the outcome. However, it should be noted that there is no consensus about the optimal number of repeated trials when evaluating reaching tasks in people with hemiparesis after stroke [155]. Fourthly, while the uFMA also includes the evaluation of reflex actions and grasp types, these types of movements cannot be assessed using the MVN Biomech system. Finally, the straps, the large number of sensors and sensor cables may have influenced the movements of participants while performing the simulated in-home task. Further developments of the body-worn sensing system are necessary before it can be used unobtrusively for evaluation of improvement or deterioration of arm movements over longer periods of time during daily life [152]. Such a system must have small-embedded sensors, not be directly visible for others, not be stigmatizing and have no influence on normal daily life behavior [9]. Many studies have described methods that could potentially be used to evaluate arm movement in daily life settings [14, 30, 63, 93, 99, 143, 149, 155]. These methods often describe arm movements in terms of acceleration, velocity or smoothness of movement of a single segment. Our proposed method combines IMU measurements on several segments for estimating metrics describing end-point hand kinematics relative to the trunk. These metrics, which evaluate relative position data, are easy to collect and may be more easily interpreted. In addition to the proposed metrics, qualifying movement performance of the arm in a daily life setting, other metrics could be evaluated. For instance, metrics which relate orientation of the upper and lower arm to describe independent joint control during functional tasks [18, 27, 29, 93, 99] or smoothness of movements [18, 30, 99, 155]. An adequate activity monitor and classifier could give context to performed arm movements, which will allow the evaluation of arm movements using the suggested metrics on a functional level.. Acknowledgments The authors would like to thank Dirk Weenk for his assistance in data collection of this study and all study participants from Roessingh rehabilitation centre for participating in this research.. 2.

(41)

(42) CHAPTER 3. Assessment of lower arm movements using one inertial sensor. Published by IEEE as: F. B. van Meulen, B.-J. F. van Beijnum, J. H. Buurke and P. H. Veltink (2017). Assessment of Lower Arm Movements Using One Inertial Sensor. Proceedings of 15th IEEE International Conference on Rehabilitation Robotics, 1407–1412. DOI: 10.1109/ICORR.2017.8009445. 3.

(43) 32. |. Chapter 3. Abstract Reduction of the number of sensors needed to evaluate arm movements, makes a system for the assessment of human body movements more suitable for clinical practice and daily life assessments. In this study, we propose an algorithm to reconstruct lower arm orientation, velocity and position, based on a sensing system which consists of only one inertial measurement unit (IMU) to the forearm. Lower arm movements were reconstructed using a single IMU and assuming that within a measurement there are moments without arm movements. The proposed algorithm, together with a single IMU attached to the forearm, may be used to evaluate lower arm movements during clinical assessments or functional tasks. In this pilot study, reconstructed quantities were compared with an optical reference system. The limits of agreement in the magnitude of the orientation vector and the norm of the velocity vectors are respectively 4.2 deg (normalized, 5.2 percent) and. 3. 7.1 cm/s (normalized, 5.8 percent). The limit of agreement of the difference between the reconstructed positions of both sensing systems were relatively greater 7.7 cm (normalized, 16.8 percent). Key words: Inertial measurement unit, Position Estimation, Algorithm, Ambulatory Assessment, Arm Movements.

(44) Assessment of arm movements using one inertial sensor. 3.1. |. 33. Introduction. Proper arm function is essential for many activities of daily living. When arm function is reduced, performance of these activities will be limited. In particular, stroke survivors may have a reduced ability to coordinate their arm movements and experience difficulties while performing daily life activities. Intensive rehabilitation therapy is usually given to restore arm function or to compensate a lack of arm function. For the optimal guidance of this rehabilitation process, arm movements should be objectively assessed during clinical assessments and functional tasks [95, 165]. A patient-specific, objective, and qualitative assessment of arm movements during functional tasks provides information on impairment level during the functional task and/or assessment, and may demonstrate recovery of arm functioning by restoration or compensation [73]. The use of an inertial measurement unit (IMU) is a feasible method for the assessment of body movements in a daily life setting [15, 95, 129, 152]. IMUs combine accelerometers, gyroscopes, and often also magnetometers. This type of sensor can be used to evaluate quantities such as orientation, change of orientation or change of position. In contrast to the use of an optical reference systems for the evaluation of body movements, IMUs do not require an external physical reference system to estimate these quantities. This in particular makes the use of IMUs suitable for measurements in a daily life setting. For example, multiple IMUs can be used for the assessment of daily life reaching performance [94]. Methods described in this example resulted in qualitative metrics to evaluate arm movements during daily life. However, the total number of IMUs needed to estimate the described metrics (at least eight sensors) makes the system less suitable for clinical assessments and daily life practice [9]. Reducing the number of IMUs to one can make these sensing systems more suitable for the evaluation of daily life movements. Although this would make the system no longer able to evaluate interactions between body parts or movements of multiple body parts, it can still be used to evaluate movements of a body part the IMU is attached to. Systems using only one IMU, attached to the lower arm or somewher else on the body, are already commonly used in rehabilitation practice. Examples include: step counters, activity monitors, the evaluation of the smoothness of movements, the assessment of overall activity, sleep cycles, amd the evaluation of body posture [15, 76, 85, 105, 114]. However, using a single IMU and additional algorithms, other clinically valuable information may be derived for instance, the quality of arm movements. Quantities such as arm velocity, arm orientation and change of arm position could be useful to estimate metrics (e.g., reaching distance and working area) for the assessment of arm movements during functional tasks and clinical assessments [94, 95]. A major drawback of using only a single IMU is the presence of signal drift when estimating velocity or the change of position of the IMU. This is inherent to using IMUs for velocity and position estimation, in which errors increase rapidly after a few seconds of measuring [47]. This study aimed to develop and evaluate a data processing method. 3.

(45) 34. |. Chapter 3. for estimation of lower arm velocity and position using a single IMU. This method could potentially be used for the assessment of arm movements during a functional task or to perform instrumented versions of already existing clinical assessments of arm function. The new method presented in this paper was developed by analogy to methods used for the reconstruction of feet movements using IMUs [134, 139, 165]. Within the methods used to reconstruct foot movements, episodes without foot movement were detected and acceleration and velocity signals of these episodes were updated. In this paper it is assumed that during a measurement of arm movements, stroke survivors are seated and there are detectable episodes without movements. The potential limitations of these assumptions are discussed.. 3.2. 3. 3.2.1. Materials and Methods System setup. In this study, one Xsens MTw Awinda IMU sensor was used (Xsens Technologies B.V., Enschede, the Netherlands). The sensor is attached using an elastic strap to the posterior side of the right forearm, just proximal of the wrist joint. The sensor is positioned along the forearm, the x-axis of the sensor frame (φs ) parallel to the forearm and pointing towards the elbow, the y-axis pointing towards the medial side of the forearm and the z-axis perpendicular to the x- and y-axis in a right-handed fashion (Fig. 3.1). Sensor acceleration, sensor angular velocity and magnetic field are internally measured at 1000 Hz. Xsens’ sensor fusion algorithms estimate the sensor orientation of the sensor frame relative to the global frame (Rgs ) [135]. All sensor data (including accelerations, angular velocity and sensor orientation) is transmitted wirelessly to a computer and collected with a sample frequency of 100 Hz.. y z. ϕs x. Figure 3.1: Overview of system setup. Xsens MTw Awinda attached to the posterior side of the right forearm, close to the wrist. φs is the sensor coordinate frame..

(46) Assessment of arm movements using one inertial sensor 3.2.2. |. 35. Sensor orientation, velocity and position estimation. The position, velocity and orientation data of the IMU in a global frame (φg , explained in more detail in Fig. 3.4), were estimated offline. All data were processed and analyzed using MATLAB® (version 2013b, MathWorks Inc., Natick, MA). To reduce noise, measured sensor accelerations and angular velocities were filtered using an eighth order Butterworth low-pass filter with a cut-off frequency of 20 Hz. Sensor velocity and sensor position were estimated using sensor acceleration and sensor orientation signals (Fig. 3.2). Noise reduction and integration methods are based on Schepers et al. [134].. mag •. gyr •. 𝑅 𝑔𝑠. SF. 𝑎𝑔. 𝑎𝑠. acc •. 𝑣𝑔. 𝑎𝑔. 𝑣𝑔. 3. 𝑥𝑔. 𝑔. Figure 3.2: Schematic overview of data processing method to estimate sensor acceleration (ag ), sensor velocity (v g ) and sensor position (xg ) in the global frame. Signals measured by the IMU (inside grey frame) are: mag (magnetometer), gyr (gyroscope) and acc (accelerometer). SF = Sensor Fusion algorithms. Rgs = sensor orientation in global frame. g = gravitational acceleration. as = sensor acceleration in sensor frame. First, measured accelerations were converted from accelerations in a sensor frame (as ), towards sensor accelerations in a global frame (ag ), by rotating the measured accelerations using the orientation of the sensor and subtracting the gravitational acceleration (g) from the z-component of the acceleration signal:. . . agx (t). . . asx (t).   0.  g      gs ay (t) = R (t) ∗ asy (t) − 0 agz (t). asz (t). (3.1). g. By integrating the sensor acceleration in a global frame, the sensor velocity in the global frame (v g ) could be estimated using: v g (t) = v g (t − 1) + ag (t) ∗ Ts. (3.2). The estimated sensor velocity is equal to the velocity of the previous sample plus the instantaneous acceleration multiplied by the sample time (Ts , the inverse sample frequency). It was assumed that the sensor velocity was zero at the first sample, v(0) = 0. By integrating the sensor velocity, the sensor position in the global frame (xg ) could.

No results found