Technology-based measurement of arm-hand skill performance in daily life conditions

Dr. R.J.M. Lemmens
Maastricht University; Adelante Centre of Expertise in Rehabilitation and Audiology, Hoensbroek
Prof. Dr. R.J.E.M. Smeets; Dr. H.A.M. Seelen; Dr. Y.J.M. Janssen-Potten

Patients with impairments of the central nervous system caused by e.g. stroke or cerebral palsy (CP) often experience serious problems with arm-hand use in daily life. Loss of arm-hand function - for example, due to a hemiparesis - limits the execution of activities of daily living (ADL) or arm-hand activities, resulting in greater dependency, restricted social participation and decreased quality of life. During rehabilitation, assessment of arm-hand skill performance is important for both clinical practice and research. Many instruments exist to assess arm-hand use in several domains of the International Classification of Functioning, Disability and Health (ICF). For instance, measuring a patient’s arm-hand function, measuring his capacity to perform activities or gauging (self-)perceived performance. However, instruments to assess actual performance in daily life are scarce. The ultimate aim of the research described in the doctoral thesis entitled ‘Technology-based measurement of arm-hand skill performance in daily life conditions’ is to develop and test an instrument to assess actual arm-hand skill performance. The instrument should be capable of measuring a) which activities a patient actually performs with his arms and hands, b) the quantity of executing (specific) activities, i.e. amount of use, and c) the quality of arm-hand skill performance. The first step was to perform a systematic literature review to give an overview and evaluate the measurements instruments available to assess arm-hand skill performance, i.e. arm-hand functioning at ICF activity level, in stroke patients and children with CP. One of the conclusions of this study was that no consensus exists about the use of instruments to assess arm-hand use on ICF activity level. Furthermore, although many instruments measure at ICF activity level, there is a lack of instruments assessing actual arm-hand skill performance. The next step was to test the currently available accelerometers as outcome measure for actual performance in a clinical study. In this randomized controlled trial (RCT), patients in the chronic phase after stroke received task-oriented robot-supported arm-hand training (intervention group) or task-oriented unsupported arm-hand training (control group). Accelerometers were used to measure the amount of arm-hand use. Whereas patients improved considerably on clinical scales and subjectively reported that they experienced improvements of their arm-hand use at home, no improvements were found in amount of arm-hand use as measured with the accelerometers. The main conclusion was that accelerometers, as used in this RCT, is not capable of discriminating between non-skill-related movements and specific skill-related movements. This is, however, essential when assessing arm-hand skill performance. A combination of multiple body-worn sensors, including accelerometers, gyroscopes and magnetometers, seems to be very promising to assess actual arm-hand skill performance. Before the instrument can be used in daily life situations, it should be tested in a standardized test situation with a limited set of activities. We have inventoried the training goals of children with CP, to be able to choose a set of activities important for patients. For each of the training goals, movement components, such as positioning the upper extremity, reach, grasp, hold, release, manipulate, push/pull/shove, displace/lift and fixate were inventoried. Based on the inventory of training goals, the activities ‘drinking from a cup’, ‘eating with knife and fork’, ‘combing hair’ and ‘opening and closing a zipper’ were selected to test the new instrument. For the studies reported in this thesis, measurements where performed with healthy adults, healthy children, adult patients with stroke and children with CP. For these measurements, sensor devices where attached to the hands, wrists, upper arms and chest of the participants. By using 7 sensor devices (3 on each arm, 1 on the chest), the combination of sensor devices resulted in 63 signals (i.e. 7 devices x 3 sensors per device (accelerometer, gyroscope and magnetometer) x three directions). Every signal contains a specific signal pattern associated with the activity performed. A prerequisite for an actual performance measure to be applied in the clinic is that its registrations of ADL can be made reliably. As a first step in testing the new instrument, the reliability of the new system was determined in healthy adults and healthy children performing activities of daily living in standardized conditions. Overall, within-subject reliability and between-subject reliability was good to very good for all activities performed by healthy adults and healthy children. The main conclusion of this study was that the instrument can reliably record ADL. Thus the instrument as used in the current setting can be used to measure skilled arm-hand performance in healthy individuals reliably. One of the purposes of the newly developed instrument is to recognize activities in registrations containing multiple different activities. In one of the studies of this thesis, the techniques to identify upper extremity activities were elucidated and proof-of-principle was shown using a set of activities tested in a healthy adult and in a stroke patient. The identification of activities is based on a pattern recognition approach called ‘template matching’. This template matching method is based on the recognition of signal patterns of a specific template, representing a specific activity, in a registration containing the performance of multiple activities. In the present thesis, the templates consisted of a combination of multiple signals, each derived from different sensors and describing the same movement from a different perspective (i.e. acceleration versus angular velocity versus orientation). Concerning the origin of the signals, both individual templates (containing signals from a registration of 1 person) and generic templates (consisting of a mean signal of 29 persons) were used for the pattern recognition. Regarding the signal content, both templates representing a complete activity and templates representing sub phases of an activity were used. For all templates, the activities drinking, eating and combing were all unambiguously recognized in registrations containing the performance of multiple activities, using data of a healthy participant as well as using data of a stroke patient. Additionally, to provide an example of the applicability in daily life, the skill identification was performed for the activity drinking in a 30 minutes daily life registration in the home setting of a healthy participant. The activity drinking was identified correctly in all cases the activity was performed. It can be concluded that this method is very promising with regard to the applicability of these sensor devices to identify specific activities among multiple skills, both in healthy individuals and patients, and both in a standardized setting and in daily life conditions. In the near future, the instrument should be optimized and tested extensively in healthy participants and patients, both adults and children. Additionally, quantification of quality of performance should be investigated and the use of the instrument in daily life conditions. Also, optimization of the instrument with regard to the design, functionality and usability should be accomplished. In the medium-term future, the research focus should lie on the implementation of the instrument in clinical practice. For the patient populations described in the present thesis as well as for other patient populations. A library can be created containing the templates of multiple persons, arranged according to diagnosis and severity of arm-hand problems. Possibilities of expanding the instrument with features to give feedback to patients about their arm-hand performance, and the application of continuous therapy should be investigated. Furthermore, the mechanisms by which deterioration of arm-hand performance, in daily life after rehabilitation treatment, occurs and by which we can better predict and identify people at risk and develop preventive treatments, should be investigated.

If you are interested in receiving a (digital) version of the PhD thesis, please contact Ryanne Lemmens (