Background: Walking disabilities negatively affect inclusion in society and quality of life and increase the risk for secondary complications. It has been shown that external feedback applied by therapists and/or robotic training devices enables individuals with gait abnormalities to consciously normalize their gait pattern. However, little is known about the effects of a technically-assisted over ground feedback therapy. The aim of this study was to assess whether automatic real-time feedback provided by a shoe-mounted inertial-sensor-based gait therapy system is feasible in individuals with gait impairments after incomplete spinal cord injury (iSCI), stroke and in the elderly. Methods: In a non-controlled proof-of-concept study, feedback by tablet computer-generated verbalized instructions was given to individuals with iSCI, stroke and old age for normalization of an individually selected gait parameter (stride length, stance or swing duration, or foot-to-ground angle). The training phase consisted of 3 consecutive visits. Four weeks post training a follow-up visit was performed. Visits started with an initial gait analysis (iGA) without feedback, followed by 5 feedback training sessions of 2–3 min and a gait analysis at the end. A universal evaluation and FB scheme based on equidistant levels of deviations from the mean normal value (1 level = 1 standard deviation (SD) of the physiological reference for the feedback parameter) was used for assessment of gait quality as well as for automated adaptation of training difficulty. Overall changes in level over iGAs were detected using a Friedman’s Test. Post-hoc testing was achieved with paired Wilcoxon Tests. The users’ satisfaction was assessed by a customized questionnaire. Results: Fifteen individuals with iSCI, 11 after stroke and 15 elderly completed the training. The average level at iGA significantly decreased over the visits in all groups (Friedman’stest, p < 0.0001), with the biggest decrease between the first and second training visit (4.78 ± 2.84 to 3.02 ± 2.43, p < 0.0001, paired Wilcoxon test). Overall, users rated the system’s usability and its therapeutic effect as positive. Conclusions: Mobile, real-time, verbalized feedback is feasible and results in a normalization of the feedback gait parameter. The results form a first basis for using real-time feedback in task-specific motor rehabilitation programs. Trial registration: DRKS00011853, retrospectively registered on 2017/03/23. Keywords: Feedback, Gait, Rehabilitation, Therapy, Incomplete spinal cord injury, Stroke, Elderly, Inertial measurement units, Wearable sensor * Correspondence: Ruediger.Rupp@med.uni-heidelberg.de Spinal Cord Injury Center, Heidelberg University Hospital, 69118 Heidelberg, Germany Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Schließmann et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:44 Page 2 of 15 Background sensors are commonly used for offline sensing / tracking Walking disabilities negatively affect inclusion in society [27, 28]. Some studies applied these sensors for balance and quality of life [1–3]. Furthermore, a non-physiological training [29–31] or gait symmetry training . Only a gait pattern increases the risk of joint overuse resulting in few mobile systems provide real-time feedback for use in pain  and - in the long run - osteoarthritis . These a therapeutic setting . Most of these systems are de- secondary complications contribute additionally to the vi- signed for providing feedback of a single gait parameter cious cycle of walking disabilities and reduction in physical specific to a certain patient group such as Parkinson’s activity, which results in an increased risk for cardiovascu- disease  or knee osteoarthritis [31, 35, 36]. Recently, lar diseases . Walking restrictions are present in pa- shoe-mounted inertial sensors were applied to estimate tients with injuries of the central nervous system, such as and increase gait quality in aged adults by measuring an- stroke  or incomplete spinal cord injury (iSCI) [8, 9] gular velocity and giving binary real-time feedback . and the associated sensorimotor and specifically proprio- So far, to our knowledge no study has provided ceptive impairments, or in aged individuals . Several proof-of-concept of an IMU-based real-time feedback factors of rather diffuse origin contribute to walking dis- training on the normalization of selected gait parameters abilities in the elderly, among them various neurological during overground walking in persons with gait disor- and orthopedic disease conditions . Similar to stroke ders of various etiologies. and iSCI, sensory impairments have a negative impact on Therefore, the aim of this pre-post-intervention, balance and walking ability [12, 13]. As afferent feedback proof-of-concept, prospective cohort study was to assess is important for motor control in a variable environment whether automatic real-time verbalized feedback pro- aswell asfor motor learning , extrinsic feedback vided by a shoe-mounted IMU-based gait therapy sys- (feedback from an external source, referred to as “feed- tem is feasible in individuals with gait impairments after back” in the following) is implemented in current gait re- iSCI, stroke and in the elderly. We hypothesized that habilitation regimes . Feedback provides an individual with this feedback system the intentional normalization with the ability to consciously close the sensorimotor con- of an individually selected gait parameter is possible. trol loop and motivates to focus on the task thereby in- creasing compliance . Material and methods In a feasibility study using a treadmill-based, real-time, Feedback implementation optical motion analysis system, individuals with iSCI and The IMU-based gait analysis and feedback system reduced knee flexion during swing phase (“stiff-knee RehaGait (HASOMED GmbH, Magdeburg, Germany) gait” like walking pattern) were able to normalize their was used for this study . It consists of a pair of IMUs gait pattern when abstract visual feedback of a weighted mounted to the user’s shoes by straps placing the distance to the physiological reference was presented sensors laterally just below the ankle joint (Fig. 1). The . Surprisingly, the study participants maintained this sensors (dimensions: 60 × 15 × 35 mm) contain a 3-axis effect even without feedback after the therapy . accelerometer (± 16 g), a gyroscope (± 2000 °/s) and a While little is known about the long-term effect of feed- 3-axis magnetic compass (± 1.3 Gs) , which was not back therapies in patients with neurological gait disor- used in the study due to large artifacts in indoor applica- ders, it has been shown in patients without neurological tions. The sensors connect to a tablet computer impairments, e.g. with patellofemoral pain syndrome, (Samsung GT-P3110, Android version 4.1.2) via Blue- that function and medical complications such as pain tooth, where the stride length, angle between foot and improve with feedback-initiated normalization of the ground at initial contact (referred to as “foot-to-ground knee joint kinematics . angle”), as well as stance and swing time were calcu- Optical, marker-based motion capture systems in com- lated with a latency of approx. 50 ms. The physio- bination with a treadmill are considered the gold stand- logical norm of these spatio-temporal gait parameters ard for instrumented gait analysis and constitute an is derived from a set of 1860 averaged gait analyses excellent platform for studies involving feedback, as they of healthy individuals (age range 5–100 years; 941 fe- provide accurate , objective movement data , and males, body height 1.61 ± 0.13 m; 919 males, body a standardized setup concerning e.g. walking speed or height 1.71 ± 0.17 m) . number of steps. However, preparation times are high A major challenge for the development of a mobile and gait kinetics on a treadmill are different from walk- feedback system was the definition of a universally applic- ing on even ground . Inertial measurement unit able evaluation scheme for classification and quantifica- (IMU)-based sensor systems represent inexpensive op- tion of the deviation from a physiological gait pattern. tions compared to marker-based gait analysis systems This evaluation scheme has to fulfill many requirements, and provide an easy-to-set-up possibility for feedback among them a feedback-parameter-independent applic- training in a natural environment [24–26]. Wearable ability, the ability for initial categorization of individuals Schließmann et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:44 Page 3 of 15 Fig. 1 Mobile, inertial measurement unit (IMU)-based gait analysis and feedback system “RehaGait” (Hasomed GmbH, Magdeburg, Germany). Shoe mounted IMUs connect to a tablet computer via Bluetooth, where gait parameters are calculated and compared to their physiological speed-corrected reference in real-time. On the basis of this calculation, automated verbalized feedback is then presented to the user independently of the underlying pathology and extent of Feedback about the distance from the normal refer- the gait disorder, the automated adaptation of training dif- ence value is categorized into three different ranges: ficulty during feedback training and the possibility to Green (correct), yellow (moderate directional instruc- evaluate the course of the training. tions) and red (strong directional instructions) (Fig. 2). For this study, a dimension-less, pathology-independent The boundaries of these ranges are defined based on the norm-distance measure for quantification of the deviation results of the initial gait analysis (iGA), which is per- from a normal gait pattern was implemented. This meas- formed at the beginning of each training visit. The start- ure integrates both the difference between the mean of a ing level is set based on the individual’s mean of the dedicated feedback parameter from an individual and an feedback parameter obtained during the iGA (Fig. 2). age- and body size-adapted norm, as well as the variance The upper boundary of the “correct”-range - the so of this norm. Based on this norm-distance measure a level called “level aim” - is set by the level boundary below structure was defined, following best practice in game de- the mean iGA value. The extent of the yellow range sign especially in serious games [40, 41]. Only with the equals to twice the SD of the feedback parameter ob- introduction of this generally applicable level structure, a tained during the iGA. The red feedback range begins at pooled data analysis of the three heterogeneous patient parameter values above the level aim plus twice the groups was possible. iGA-SD and has no upper boundary (Fig. 2). During the In detail, the range for each level of the norm-distance training session, after every third stride, a feedback value measure is centered around the mean value of the is calculated by comparing the median of the preceding physiological norm of the respective feedback training 3 strides to the current level aim. To feed back this value parameter and has the size of one standard deviation 3 basic feedback instructions are used according to the (SD) of the norm representing the physiological “noise” above-mentioned categories, e.g., “stance duration much (Fig. 2). Levels start from 0 with no upper limit. Level 0 shorter” (red), “stance duration shorter” (yellow) and represents the lowest level with the least deviation from “stance duration correct” (green) (Fig. 2). Feedback in- the norm, while a level n indicates that the current structions are generated by the tablet computer using deviation of the mean value ranges between n and n + 1 the Android-OS default Google text-to-speech engine. SDs of the norm. Thus, the level structure forms a posi- Every 20 strides, the program logic decides on the tive linear scale for unambiguous assignment of a devi- adaption of the training level: If more than 70% of the ation from the norm to a dedicated level. With this feedback values fell into the red area the level is in- categorization principle, each individual can be assigned creased, i.e., the difficulty to achieve the level aim is de- a level independent of the gait abnormality. creased. On the contrary, if > 70% fell into the green Schließmann et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:44 Page 4 of 15 Fig. 2 Level structure implemented in the mobile feedback system and exemplary course of levels over training. Mean and standard deviation (SD) of the respective physiological reference (dark red) constitute the absolute training aim and the distance between training levels, respectively. In an initial gait analysis (GA), mean and SD of the user are calculated and used to define the training level at therapy onset together with the level aim and the width of the yellow feedback area (directional feedback). During training, levels change depending on the percentage of feedback values in the red and green range of values, thereby automatically adjusting training difficulty. Training difficulty increases (level decreases) if more than 70% ofthe feedback values are in the green range over a defined number of strides (20 in this study). In this example, the initial level at training onset was 3 and changed to 2, decreased to 1 and then went up to 2 again. The feedback structure below the physiological mean, graphically represented by the shaded feedback areas, is fixed and follows the structure of level 0. For a better understanding, a situation is depicted in which the mean of the feedback parameter obtained in the initial GA of the patient is higher than the physiological mean. In the opposite situation the feedback structure would be mirrored at the level of the physiological mean area, the level is decreased, i.e., the difficulty to achieve Scale (AIS) grade C or D ) were included not earlier the level aim is increased (Fig. 2). If neither of these two than 6 weeks after onset of paralysis. Individuals with conditions apply, the level remains unchanged for an- ischemic or hemorrhagic stroke were included not earlier other 20 strides. To prevent overcorrection, the feedback than 3 months after onset of paresis. A Mini structure past the physiological mean is not changing Mental-Status-Test  was applied in individuals with and always follows the structure of level 0. To prevent stroke to screen for cognitive impairments. Individuals guidance effects and encourage self-reflection , a with severe dementia (Mini Mental-Status-Test total fading feedback approach  was implemented: score < 18) were not included. Individuals with iSCI or In case a user of the system fulfilled the current move- stroke were excluded in case of strong spasticity (Modified ment aim, the frequency of positive feedback was step- Ashworth Scale (MAS) , unilateral score > = 10) or a wise reduced (feedback output every 3rd, 6th, 15th, 30th, history of epileptic seizures. Elderly with an age of at least 45th, etc. stride), until directional feedback (red or yel- 65 years and not specifically defined gait abnormalities low category) was given. Directional feedback was always were included. Subjects with a history of a previous stroke provided every 3rd stride. or Parkinson’s disease (freezing and/or shuffling gait ham- per gait phase detection with RehaGait) were excluded Inclusion/exclusion criteria from the study. Study participants had to perform Legal age, a gait abnormality caused by a stroke or complete foot clearance during swing phase and a incomplete spinal cord injury (iSCI) or confirmed by an foot-flat phase (both heel and toes are on the ground ) observational gait analysis of a clinical specialist (elderly), during stance, because these are crucial for gait event de- the ability to walk for 20 min and at least 100 m in the tection by the RehaGait system. A sample size of 15 indi- 6-min walk test  were the common inclusion criteria viduals per group was intended for feasibility testing. An for all groups. In general, all walking aids were allowed, a-priori sample size calculation was not possible as no ef- but only ankle-foot-orthoses were allowed as braces. fect size of this novel therapy method is known. Individuals with iSCI at any neurological level of injury Individuals with iSCI and stroke were recruited from and preserved motor function below the lesion level the inpatient or outpatient service at the Spinal Cord (American Spinal Injury Association (ASIA) Impairment Injury Center in Heidelberg, Germany. For outpatient Schließmann et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:44 Page 5 of 15 recruitment an outpatient call for participation was pub- lished. Elderly were recruited in the day care unit of the geriatric hospital of the University of Jena, Germany, during an inpatient stay for general check-up. The recruiting period extended from July 2015 to November 2016. In all individuals, no expense reimbursement was provided for participation. All study participants were instructed to continue their therapy programs over the course of the study. Prior to study inclusion, written informed consent was obtained from all participants. Ethical approval was granted by the ethics committees of the medical faculty of Heidelberg University (S-168/2015) and Jena Univer- sity (4377–04/15). The study has been registered with ID DRKS00011853 in the German Register for Clinical Trials (DRKS). The study protocol remained unchanged over the whole runtime of the study. Clinical assessments and users’ satisfaction survey Fig. 3 Overview of the study protocol. Gait analyses (yellow, GA) In the iSCI group, neurological impairment was assessed over 25 strides are performed before and after the training, as well with the International Standards for Neurological Classi- as at the follow-up visit. Training sessions (blue) are interrupted by fication of Spinal Cord Injury (ISNCSCI) . The pauses (P) of no more than 2 min each. During gait analyses and walking tests (10-m walk test (10MWT) and Timed up and go test ISNCSCI assessment involves functional testing of 5 key (TUG)) no feedback (FB) is given muscles in each of the upper and lower extremities as well as light touch and pin-prick testing in 28 derma- tomes, to estimate location and severity of the SCI. Par- for the iSCI and stroke groups only. A follow-up assess- ticipants with stroke underwent a manual muscle test of ment of elderly individuals was not possible due to a key muscles (flexors and extensors of hip, knee and limited stay (< 3 weeks) in the outpatient clinic of the ankle, scores 0–5 as defined in ISNCSCI, referred to as University of Jena. The 10MWT and TUG were “Mscores”) as well as an ISNCSCI compatible pin-prick conducted before and after the training and during the testing. For reasons of comparability, the WISCI II  follow-up visit. Walking assessments and feedback was also applied in the stroke group to describe the de- training were carried out indoors on a barrier-free, pendency on walking aids. To test for potential changes even-ground walkway by one assessor in the respective in walking ability, the Timed up and go test (TUG)  center. and the 10-m walk test (10MWT)  were used. The end user’s satisfaction with the therapy system was ob- Training procedure tained post training by a questionnaire based on the In each individual, the gait parameter with the highest Quebec user evaluation of satisfaction with assistive level, i.e. with the highest deviation from the physio- technology (QUEST) . logical norm, obtained in the first iGA was chosen as feedback parameter. In case individuals were not able to Study protocol voluntarily influence the chosen parameter in the resting The protocol consisted of 3 consecutive training visits position, an alternative feedback parameter was chosen. and 1 follow-up assessment 4 weeks post training (Fig. 3). Prior to the first feedback therapy session, individuals Each training visit started with an iGA over 25 strides, were informed about the intent and frequency of the followed by 5 sessions of feedback training over 3 min. verbalized feedback, and the meaning of a level and its The first iGA of the first training visit served as the ini- dynamic change. tial baseline for level calculation . For patients who All walking related assessments were conducted reported not being able to complete a 3-min-session using the same shoes and assistive devices. Assess- and/or reported mental or physical fatigue, training ses- ments were interrupted if individuals reported an in- sion duration could be reduced to 2 or 1.5 min based on creasing fear to fall or if exhaustion jeopardized to the therapists’ decision. The minimum total therapy dur- safely continue the intervention. For safety reasons, ation for data evaluation was 22.5 min. Each training an assistant walked next to the study participants. visit ended with a post gait analysis (pGA). A follow-up Participants were free to choose earphones or to lis- assessment was made four weeks after the training phase ten to the build-in speakers of the tablet computer. Schließmann et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:44 Page 6 of 15 Throughout all therapy sessions, study participants were free to choose their comfortable walking speed and the strategy to accomplish the task imposed by the verbalized feedback instructions. Data evaluation & statistics Data from participants who completed the 3 days of training were evaluated. Main outcome parameter was the average level of the trained parameter of iGAs on visits 1 (baseline) through 3 calculated per individual over 25 measured strides, while missing values for single strides were omitted. A Friedman’s test was used to de- tect overall significance. Bonferroni-corrected Wilcoxon post-hoc tests were used to detect changes between training visits. Since non-parametric tests were used, data is reported by median and 25th and 75th percen- tiles (median (25, 75%)). For analysis of the overall thera- peutic effect additional evaluations were made by calculating the average level per iGA of all levels of pa- rameters not trained (including contralateral body side) in the respective individual. In order to get an impression about changes in actual values of gait parameters not in the focus of the feed- back training (contralateral side and other time-distance Fig. 4 Recruitment process flowchart. Altogether, 41 individuals finished parameters), a subgroup consisting of individuals train- the training period of the study and their data was analyzed. A follow-up ing foot-to-ground angle, the most frequently chosen was not planned in the elderly group. Colors for patient groups parameter, was made. These gait parameters (stride are consistent in all graphs length, foot-to-ground angle, stance duration, swing dur- ation) were evaluated over the first 3 iGAs. Walking tests (TUG, 10MWT) were carried out before iGA 1, after pGA 3 and after follow-up (Fig. 3). Gait parameters 132 months post injury) with a lower extremity motor of the subgroup and walking tests were evaluated using score of 40.7 ± 6.7 (max. 50) (Table 1). Stroke individuals paired Wilcoxon tests. (4f, 7 m, 57 ± 8 years, 8 ischemic, 2 hemorrhagic, one End user questionnaires were evaluated with descrip- unresolved cause) were included 66 ± 74 months after tive statistics using boxplots and histograms. onset of paralysis. All stroke individuals suffered from Statistical evaluations were performed using R version hemiparesis (Mscores non-impaired body side: 29.89 ± 3.2.1 . The threshold for significance was α < 0.05. 0.33 (max. 30), Mscores impaired body side: 21.78 ± Graphics were generated using the ggplot2 R package 4.35; Table 1). Elderly participants (7f, 8 m) were 81 ±  and Inkscape version 0.91. 6 years old. The mean Mini Mental-Status-Test total score was 29.4 ± 0.8 (range 28–30). Results Participants Fifty-six individuals (19 iSCI, 14 stroke and 23 elderly in- Training parameters and training time dividuals) were screened for eligibility, 7 of which could Foot-to-ground angle was the most frequently used not be included due to insufficient walking abilities. training parameter in all groups. (Fig. 5). Total average Forty-one individuals (15 iSCI, 11 stroke and 15 elderly therapy time was 36.3 ± 8.9 min for individuals with iSCI individuals) participated in all training visits and were and 37.5 ± 7.4 min for individuals after stroke, respect- included in the analysis. Eight elderly subjects dropped ively. In both patient groups, each participant completed out due to an insufficient number of training sessions (< 5 training sessions per training visit. Therapy time for 5 training sessions/visit) or voluntary withdrawal of con- elderly individuals was > 35 ± 4.3 min with a training sent. Six individuals (4 iSCI, 2 stroke) were lost to session time between 2.5 min and 3 min. Four elderly follow-up (Fig. 4). Individuals with iSCI (11f, 4 m, 53 ± individuals did not complete all 5 training sessions per 18 years, 6 cervical, 6 thoracic, 3 lumbar) were all classi- training visit, but complied with the minimally required fied as AIS D at the time point of inclusion (57 ± total therapy duration. Schließmann et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:44 Page 7 of 15 Table 1 Patient description group/ID sex age MAI WISCI II 6-Min-Test LEMS  PP  AIS NLI iSCI1 f 29 24 20 494 46 101 D L1 iSCI2 f 52 120 13 150.5 31 33 D C2 iSCI3 f 62 35 20 353 43 81 D T5 iSCI4 f 49 6 20 331.5 48 95 D C4 iSCI5 m 61 64 20 445 46 85 D T6 iSCI6 m 59 4 18 171 41 56 D C5 iSCI7 f 57 34 13 100 26 68 D T5 iSCI8 m 72 3 19 359 44 37 D C2 iSCI9 m 46 3 13 301.5 45 95 D L1 iSCI10 f 57 1 13 338 40 74 D T10 iSCI11 f 69 2 9 215 47 D C4 iSCI12 f 18 13 16 307 41 62 D T5 iSCI13 f 59 520 18 319.5 90 D T12 iSCI14 f 77 27 13 299.5 41 42 D C3 iSCI15 f 20 2 16 245 31 100 D L2 mean 52.5 57.2 16.0 295.3 40.7 72.8 SD 17.7 132.0 13; 18.5 105.8 6.7 23.7 Mscore  PP  hemisphere - artery stroke1 f 57 12 20 283 19 23 right stroke2 m 53 127 20 500 left - A. cerebri media stroke3 m 39 247 15 334.5 24 56 right - A. cerebri media stroke4 f 72 9 20 500 27 44 right - A. cerebri media stroke5 m 58 65 20 444.5 23 55 left - A. basilis stroke6 m 59 23 15 219.5 12 0 right - A. carotis stroke7 f 58 33 20 294 23 42 left - A. carotis interna stroke8 m 58 130 18 347.5 24 52 right - A. carotis stroke9 m 54 39 15 277 20 47 left - A. vertebralis stroke10 f 56 31 15 344.5 left - A. cerebri media stroke11 m 61 9 20 378 24 54 left - A. cerebri media mean 56.8 65.9 20.0 356.6 21.8 41.4 SD 7.7 74.1 15; 20 91.9 4.4 18.6 elderly1 f 82 elderly2 m 76 elderly3 f 73 elderly4 m 89 elderly5 f 72 elderly6 m 77 elderly7 m 90 elderly8 f 77 elderly9 m 82 elderly10 m 82 elderly11 m 90 elderly12 f 84 Schließmann et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:44 Page 8 of 15 Table 1 Patient description (Continued) group/ID sex age MAI WISCI II 6-Min-Test LEMS  PP  AIS NLI elderly13 m 78 elderly14 f 76 elderly15 f 86 mean 80.9 SD 6.0 Characteristics of individuals with incomplete spinal cord injury (iSCI), after stroke, and with old age (elderly) included in data evaluation. Sex and age are given for all groups. For the group with iSCI, months after injury (MAI), data according to the International Standards for Neurological Classification of Spinal Cord Injury (ISNCSCI), Walking Index for Spinal Cord Injury II (WISCI II), 6-min walk test (6-Min-Test), Lower Extremity Motor Score (LEMS), Pin Prick score (PP), ASIA Impairment Scale (AIS) and Neurological Level of Injury (NLI) are listed. For individuals after stroke, key muscles motor scores (flexors and extensors of hip, knee and foot, motor score grading according to MRC) as well as ISNCSCI pin-prick testing have been determined (Mscore and PP, respectively) and are listed for the affected body side. Maximal scores are displayed in brackets. Additionally, information about the location of the insult (brain hemisphere and artery) is listed for this group For reasons of consistency, WISCI II scores were given for individuals with stroke, too. For WISCI II scores, median and 1st and 3rd quartile replace mean and SD Changes in training levels From the pGA after training visit 3 to the iGA at A pooled analysis of all 3 groups together revealed a sig- follow-up 4 weeks post training, the iSCI group showed nificant reduction of the level (= 1 x SD of the reference a decrease of 1 level, while the stroke group’s level of the feedback parameter) over time (Friedman’s Test p stayed constant. < 0.0001, Table 2). Post-hoc tests between the 3 iGAs of The average level of all non-trained gait parameters all training visits revealed a significant decrease of the of both body sides did not change significantly over the iGA level from visit 1 to visit 2 (median level 5 (3, 6) iGAs, if data of all patient groups were pooled (initial and 3 (2, 4)), and from visit 1 to visit 3 (median level 2 1.67 (1.17, 3), 1.5 (1, 3) at iGA2, 1.67 (1, 2.67) at iGA3). (1, 4)), respectively. While individuals with iSCI and the elderly started at All groups had a high initial level (iSCI: 4 (3, 5), stroke moderate values in iGA1 (2 (1.09, 2.83), 2.33 (1.5, 3.83), 5 (2, 6), elderly 6 (4, 6)) and experienced the highest re- respectively), stroke individuals showed low initial duction in levels on the first training visit (Fig. 6 and levels (1.17 (0.92, 1.59)) in all non-trained parameters. Additional file 1). In the elderly, levels increased during A significant reduction (p = 0.0096, N =15) of the level the second and third training day (2 (0.5, 4.5) to 3 (1.5, was only found in the elderly, in whom levels between 5), p = 0.1975 and 2 (1, 5) to 3 (1, 5), p = 0.4263, respect- iGA1 and iGA2 were reduced from 2.33 (1.5, 3.83) to ively, paired Wilcoxon Test) and started with a lower 1.5 (0.915, 3.33). level on visit 2 and 3 compared to the pGA of the previ- ous visit (Fig. 6d, Additional file 1). Subgroup gait parameters Within the subgroup of individuals with iSCI training foot-to-ground angle (iSCI: 11, stroke: 5, elderly: 12), foot-to-ground angles increased significantly towards the physiological reference (p = 0.0186, N = 11) on the ipsi- lateral side (Fig. 7) between iGA1 and iGA3, as well as between iGA1 and iGA2 (p = 0.0244). On the contralat- eral body side, no significant increase was found for this group. For the subgroup of individuals with stroke train- ing foot-to-ground angle, no significant changes were found in any gait parameter. The subgroup of elderly training foot-to-ground angle normalized this parameter between iGA1 and iGA3 / iGA2 on the ipsilateral (p = 0.001, p = 0.0005, respectively) and contralateral (p = 0.0005, p = 0.0024, respectively) body side. Similarly, stride length increased towards the reference between iGA1 and iGA3 / iGA2 (p = 0.0024, p = 0.001, respectively). Fig. 5 Distribution of training parameters chosen for feedback therapy ordered by group of participants. Foot-to-ground angle is defined as Walking test results the angle between foot and ground at heel strike. Note that 2 in 11 The average initial comfortable walking speed captured stroke individuals were trained with feedback of a gait parameter on by the 10MWT before training was 0.91 (0.62, 1.02) m/s the unaffected body side (2× stance duration of the unimpaired leg) in the iSCI group, 1.06 (0.78, 1.19) m/s in the stroke Schließmann et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:44 Page 9 of 15 Table 2 Overview of the statistical results Friedman’s Test Wilcoxon post-hoc test iGA 1 < −> 2 iGA 1 < −> 3 iGA 2 < −>3 overall N =41 p < 0.0001 * p < 0.0001 * p < 0.0001 * p = 0.2450 −5 ci1 = 1.50 ci2 = 2.50 ci1 = 1.50 ci2 = 2.50 ci1 = −3.93e ci2 = 1.00 iSCI N = 15 p = 0.0004 * p = 0.0027 * p = 0.0047 * p = 0.1883 − 5 ci1 = 1.00 ci2 = 1.00 ci1 = 1.00 ci2 = 2.00 ci1 = −6.45*e ci2 = 2.00 stroke N = 11 p = 0.0019 * p = 0.0131 * p = 0.0069 * p = 0.5862 ci1 = 1.00 ci2 = 3.50 ci1 = 1.00 ci2 = 3.00 ci1 = −1.00 ci2 = 2.00 elderly N =15 p = 0.0001 * p = 0.0010 * p = 0.0016 * p =1 ci1 = 1.50 ci2 = 3.00 ci1 = 1.50 ci2 = 3.50 ci1 = −2.00 ci2 = 1.50 Overview of the statistical results of the changes of the average training levels over initial gait analyses (iGA) of visits 1 through 3, by participant group. Overall significance testing was achieved with a Friedman’s Test. For post-hoc comparisons, a Bonferroni corrected paired Wilcoxon test was used. Significant results are marked with an asterisk (*) group and 0.71 (0.52, 0.83) m/s in the group of elderly. analysis of the change in levels from iGAs 1 to 3 of the There were no significant changes in walking speed subgroups with feedback parameter “foot-to-ground measured by the 10MWT and time needed for comple- angle” (N = 28) and “time-distance parameters” (pooling tion of the TUG for either group over the course of the all remaining individuals, N = 13) suggested that the im- study including follow-up (Additional file 2). provements are not dependent on the trained feedback parameter: In the “foot-to-ground angle” subgroup the Questionnaire on user satisfaction median level improved from 4 (3, 6) (median and Participants (14 iSCI, 11 stroke and 14 elderly) reported percentiles 25 and 75) at iGA 1 to 2.5 (2, 3.25) at to be overall very satisfied with the system (Fig. 8a), iGA 2 and finally 2 (1, 4) at iGA3. Similarly, the while study participants with iSCI assigned the highest “time-distance parameters” subgroup started initially satisfaction scores. Participants were asked to choose 2 at 6(4, 8), reduced the level to 4 (1, 7) at iGA2 and out of 5 categories which they rated as most important. remained at 4 (1, 5) at iGA 3. Efficacy was chosen most frequently in all 3 groups, Similarly to our previous findings most improve- while elderly individuals also had a strong focus on ments occurred during the first training visit whereas no safety (Fig. 8b). The responses to category-specific ques- significant change was found in the two subsequent train- tions yielded overall positive results (Fig. 8c). Elderly in- ing visits. This rapid normalization of the feedback train- dividuals were only moderately satisfied with the effect ing parameter within minutes indicates that participants of feedback on the perceived changes on the gait pattern, were enabled by the automated feedback from RehaGait with the changes in training level, and with the possibil- to make full use of their preserved muscles strength and ity to accomplish long-lasting gait pattern changes. One coordination ability to accomplish the given task. This individual with a stroke associated aphasia suggested the finding is in line with clinical evaluation results from other implementation of an optional feedback output using mobile feedback systems , although a direct compari- simple sounds instead of polysyllabic words. son of results is not possible due to differences in the technical specifications of the feedback systems and vary- Discussion ing study protocols with different outcome parameters The results of our baseline proof-of-concept study dem- . Furthermore, such fast positive effects might be onstrate the feasibility of a mobile, verbalized, real-time linked to trick movements in proximal joints such as the gait kinematics feedback therapy using shoe-mounted hip or knee joint. Even though the presence of trick IMUs. Overall, the introduced level concept was found movements cannot be completely excluded, our study to be well suited as a universal measure for evaluation of participants did not show a deterioration of the gait abnormalities independent of the underlying path- non-feedback-trained gait parameters measured on both ology as well as for therapy progress monitoring inde- legs by the shoe-mounted system indicating a true pendently of the feedback parameter. The three different normalization of the overall gait pattern. The results of groups of study participants comprising gait abnormal- this proof-of-concept-study underline the fact that a suffi- ities resulting from different etiologies were able to sig- cient degree of sensory, either preserved intrinsic or - in nificantly normalize the trained feedback gait parameter case of impaired intrinsic feedback - extrinsic feedback over the course of three consecutive therapy visits. In needs to be present for proper motor control , an issue about 2/3 of the study participants, the foot-to-ground which is often not sufficiently taken into account in motor angle was used as feedback parameter. A qualitative rehabilitation strategies. Schließmann et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:44 Page 10 of 15 ab cd Fig. 6 Boxplots with the mean level at initial gait analysis and gait analysis after training (post) on each training visit and at follow-up for all participants (a), and separately for the groups with incomplete spinal cord injury (iSCI) (b), stroke (c) and the elderly (d). Diamond shapes indicate mean values. Horizontal bars indicate significant differences according to paired Wilcoxon tests The fact that changes of the feedback parameter to- feedback therapy. It might be more effective to avoid wards more physiological values occurred very rapidly mass practice regimes (e.g., 5 therapy visits in one has some implications not only for the quick identifi- week) and apply the feedback therapy over an ex- cation of non-responders, but also on the future tended period (e.g., 1 therapy visit per week over clinical implementation of the automated real-time 5 weeks) of time. Additionally, this helps to prevent a Schließmann et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:44 Page 11 of 15 ab cd Fig. 7 Boxplots showing parameter values “foot-to-ground angle” (a), “stride length” (b), “stance duration” (c) and “swing duration” (d) on initial gait analyses (iGA) 1 through 3 of a subgroup of individuals who were trained with “foot-to-ground angle” as training parameter. Ipsilateral and contralateral body sides are plotted separately (dark and light color, respectively). Diamond shapes indicate mean values dependency on the external feedback. It might also be the drawback of a higher system complexity and longer possible to consecutively train different gait parame- preparation times. ters (potentially associated with different underlying At follow-up 4 weeks after training, positive carryover causes of the gait abnormality), while monitoring for effects were still present in individuals with iSCI and increased norm deviation of the parameters not in stroke, indicating that some motor learning took place. focus of training. However, these hypotheses need to However, it is not known in how far concomitant be confirmed in future studies involving a substantial physiotherapy contributed to this process. On the number of participants. second and third visit, only in the elderly, the deviation From a technological viewpoint, the feedback system from the norm increased (although not significant) might support the selection of the feedback parameter during the instrumented feedback therapy, where study in the future by implementation of an automated selec- participants have to concentrate on the feedback instruc- tion algorithm based on a quantitative analysis of levels tions and simultaneously on the adaption of their gait of different gait parameters. Additionally, the use of pattern. This decrease in task performance is known also more sensors would allow for an assessment and feed- from other dual-task gait studies . Furthermore, back of the kinematics of multiple joints and thereby elderly might rely on the recruitment of many brain controlling for potential trick movements, however, with regions for compensatory purposes during motor tasks Schließmann et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:44 Page 12 of 15 ab Fig. 8 Results of the user’s satisfaction questionnaire (modified Quebec user evaluation of satisfaction with assistive technology (QUEST)). Fourteen individuals with incomplete spinal cord injury (iSCI), 11 with stroke and 14 elderly took part in the survey. Overall satisfaction was captured by a visual analogue scale (a), where a score of 1 depicts highest and 0 lowest satisfaction. The categories of the questionnaire rated as most important are displayed in a histogram (b). Each participant could choose up to 2 out of 5 categories. Within each category, specific questions could be answered on an ordinal scale from 1 (not satisfied at all) to 5 (very satisfied). These results are displayed using box plots; mean values are displayed using diamond shapes (c) , thus increasing mental workload and decreasing at- The analysis of the pre-post therapy walking tests re- tention. Therefore, it might be advisable to introduce vealed no improvements, although all three patient shorter training sessions in order to avoid mental and/or groups had a lower mean initial comfortable walking physical exhaustion specifically in persons with reduced speed (iSCI: 0.91 m/s, stroke: 1.06 m/s, elderly: 0.71 m/s) cognitive abilities. compared to non-disabled persons (approx. 1.4 m/s Schließmann et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:44 Page 13 of 15 for people in their 50s, approx. 1.3 m/s for people in the pathological gait pattern, which might have resulted in their 70s) . However, due to the small amount of a selection bias. only 3 training visits, we did not expect a substantial im- This analysis of the study results faced a multiple out- provement. On the other hand, study participants did come measure and multiple testing problem in an effort not show a slower walking speed, although they were to generate a maximum number of hypotheses with the trying to maintain a more physiological walking pattern limited number of study participants and assessment during the walking tests. visits. Corrections for multiple testing were limited to Our evaluation of the RehaGait real-time feedback-system post-hoc comparisons of the main outcome measure. followed a user-centered design approach obtaining feedback from the users about their satisfaction with Conclusions different aspects of the device. Despite the fact that The results of this proof-of-concept study show that a the walking tests were not able to capture the effects standardized, etiology-independent, mobile feedback of the feedback therapy on the improvement of the therapy based on shoe-mounted IMU sensors is feasible. quality of the walking pattern, the overall positive rat- Providing external real-time feedback about the deviations ings in the users’ satisfaction questionnaire in particu- from a normal gait pattern addresses an important lar on the therapeutic efficacy of the system can be problem in restorative walking rehabilitation programs interpreted as a proof for the perceived benefits of and may effectively contribute to task-oriented the therapy from the user side. The verbalization of locomotion therapies aiming at the normalization of a the feedback was well received, however given the non-physiological gait pattern. Future studies are needed substantial number of potential future users with to define the frequency and individual implementation of aphasia (20–30% of initial stroke [61, 62]), alternative feedback therapy paradigms and should log changes in hip auditory feedback, e.g., through sounds, should be and knee angles. considered for this group. In general, future studies need to determine the minimal functional require- ments of feedback therapy responders. Our study was Additional files intended as a proof-of-concept that individuals with different gait disorders including lesions of the CNS Additional file 1: Level data for all visits. Level at pre-training (pre) and post-training (post) gait analyses (GAs) on visits 1 through 3 and follow- and the associated muscular weakness are capable to up arranged by participant group. Median and percentiles 25 and 75 (in normalize a selected gait parameter by real-time braces) are listed. (DOCX 16 kb) feedback. If this normalization results in an improved Additional file 2: Results of the 10-m walk test (10MWT) and Timed up clinical outcome, e.g. reduction of pain or other and go Test (TUG). Groupwise results of the 10MWT and TUG performed before the 1st gait analysis (GA) on visit 1, after the post GA of visit 3 and musculoskeletal complications, needs to be shown in at follow-up. Median and percentiles 25 and 75 (in braces) are listed. future studies with long-term follow-up. The (DOCX 16 kb) pre-post-intervention baseline study design was Additional file 3: Complete dataset of gait parameters and levels measured possible for this proof-of-concept study, because the / calculated by RehaGait. Complete dataset listing average levels per individual per gait analysis, as well as mean gait parameters (stance and swing duration, involved individuals with a chronic gait disorder have foot-to-ground angle, stride length) measured by the RehaGait system. very limited potential for spontaneous recovery. This (XLSX 58 kb) would be different in individuals in the acute or sub- Additional file 4: Complete dataset of walk test results. Complete acute stage, where the involvement of a control group dataset listing results of the 10-m walk test (10MWT) and Timed up and go test (TUG). (XLSX 12 kb) is mandatory. Additional file 5: Complete dataset of end user survey results. Complete Future work should evaluate in how far the level struc- dataset listing results on the questionnaire based on the modified Quebec ture of the norm-distance measure introduced in this user evaluation of satisfaction with assistive technology (QUEST). Responses work can be used for objective assessment of gait disor- on a 5 point ordinal scale (1 = not satisfied at all, 2 = not satisfied, 3 = more or less satisfied, 4 = quite satisfied, 5 = very satisfied) of category-specific ders and of effects of interventions. For this, the relation- questions are followed by the selection of users’ most important categories ship between the levels and established, clinical, gait (1 = selected). The last column contains the results of a visual analogue scale disorder etiology-specific, ordinally scaled gait assessment (VAS, 0–1) rating. (XLSX 13 kb) tools such as the Spinal Cord Injury Functional Ambula- tion Inventory SCI-FAI  or the Functional Gait As- Abbreviations sessment  needs to be determined . 10MWT: 10-m walk test; AIS: ASIA Impairment Scale; ASIA: American Spinal Injury Association; DRKS: German Register for Clinical Trials; iGA: initial gait Study limitations analysis; IMU: inertial measurement unit; iSCI: incomplete spinal cord injury; ISNCSCI: International Standards for Neurological Classification of Spinal Cord To foster fast recruitment of study participants, the inclu- Injury; MAS: Modified Ashworth scale; pGA: post gait analysis; QUEST: Quebec sion criteria were very broad resulting in inclusion of all user evaluation of satisfaction with assistive technology; TUG: Timed up and available ambulatory patients without further specifying go test Schließmann et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:44 Page 14 of 15 Acknowledgements 7. Mendis S. Stroke disability and rehabilitation of stroke: World Health We thank all participants for taking part in this study. Many thanks to Dr. Organization perspective. Int J Stroke. 2013;8:3–4. Tom Bruckner from the Department of Medical Biometry of the Heidelberg 8. White N-H, Black N-H: Spinal cord injury (SCI) facts and figures at a University Hospital for his statistical support and to Dr. Kwetkat for patient glance. 2016. recruitment in the day unit of the geriatric hospital of the University of Jena. 9. McDonald JW, Sadowsky C. Spinal-cord injury. Lancet. 2002;359:417–25. We thank Marc Hofmann and Nils Brauckmann from HASOMED GmbH for 10. Bridenbaugh SA, Kressig RW. Laboratory review: the role of gait analysis in technical support. seniors’ mobility and fall prevention. Gerontology. 2010;57:256–64. 11. Sudarsky L. Gait disorders in the elderly. N Engl J Med. 1990;322:1441–6. Funding 12. Lord SR, Rogers MW, Howland A, Fitzpatrick R. Lateral stability, sensorimotor This work is funded by the German Federal Ministry of Economic Affairs and function and falls in older people. J Am Geriatr Soc. 1999;47:1077–81. Energy (BMWi), project “RehaGait”, grant number KF2906702KJ2. 13. Goble DJ, Coxon JP, Wenderoth N, Van Impe A, Swinnen SP. Proprioceptive sensibility in the elderly: degeneration, functional consequences and plastic- Availability of data and materials adaptive processes. Neurosci Biobehav Rev. 2009;33:271–8. All data generated or analyzed during this study are included in the 14. Riemann BL, Lephart SM. The sensorimotor system, part II: the role of Additional files 3, 4 and 5 in this published article. proprioception in motor control and functional joint stability. J Athl Train. 2002;37:80–4. Authors’ contributions 15. Krakauer JW. Motor learning: its relevance to stroke recovery and DS, CS, RR designed and conducted the proof-of-concept study for individuals neurorehabilitation. Curr Opin Neurol. 2006;19:84–90. with iSCI and stroke. MN, SD, US designed and conducted the proof-of-concept 16. Sigrist R,RauterG,Riener R, Wolf P.Augmented visual, auditory, haptic, and study for elderly individuals. DS, RR, NW, MN, SD were responsible for fulfilling multimodal feedback in motor learning: a review. Psychon Bull Rev. 2013;20:21–53. the legal requirements for the conduct of the study. DS with the support of CS 17. Colombo R, Pisano F, Mazzone A, Delconte C, Micera S, Carrozza MC, Dario P, performed the processing of all clinical and instrumented data and did the Minuco G. Design strategies to improve patient motivation during robot-aided statistical evaluation. All authors were involved in the development of rehabilitation. Journal of neuroengineering and rehabilitation. 2007;4:3. the level structure and of the general feedback therapy concept. All 18. Wolf S, Loose T, Schablowski M, Doderlein L, Rupp R, Gerner HJ, Bretthauer authors contributed to the writing and approved the final version of the G, Mikut R. Automated feature assessment in instrumented gait analysis. manuscript, and are therefore liable for its content. Gait Posture. 2006;23:331–8. 19. Schlieβmann D, Schuld C, Schneiders M, Derlien S, Glöckner M, Gladow T, Ethics approval and consent to participate Weidner N, Rupp R. Feasibility of visual instrumented movement feedback therapy Ethical approval was granted by the ethics committees of the medical in individuals with motor incomplete spinal cord injury walking on a treadmill. faculty of Heidelberg University (S-168/2015) and Jena University (4377–04/15). Front Hum Neurosci. 2014;8:416. https://doi.org/10.3389/fnhum.2014.00416. Prior to study inclusion, written informed consent was obtained from all 20. Noehren B, Scholz J, Davis I. The effect of real-time gait retraining on hip participants. kinematics, pain and function in subjects with patellofemoral pain syndrome. Br J Sports Med. 2010;45(9):691–6. Competing interests 21. Carse B, Meadows B, Bowers R, Rowe P. Affordable clinical gait analysis: an HASOMED GmbH, 39,114 Magdeburg, Germany was involved in development assessment of the marker tracking accuracy of a new low-cost optical 3D of the system and provided the hardware and software used in this motion analysis system. Physiotherapy. 2013;99:347–51. study. HASOMED GmbH was not involved in development of the study 22. Coutts F. Gait analysis in the therapeutic environment. Man Ther. 1999;4:2–10. design, conduction of the study or any data evaluation. We certify that 23. Watt JR, Franz JR, Jackson K, Dicharry J, Riley PO, Kerrigan DC. A three- no party having a direct interest in the results of the research supporting this dimensional kinematic and kinetic comparison of overground and treadmill article has or will confer a benefit on us or on any organization with which we walking in healthy elderly subjects. Clin Biomech. 2010;25:444–9. are associated. 24. Mayagoitia RE, Nene AV, Veltink PH. Accelerometer and rate gyroscope measurement of kinematics: an inexpensive alternative to optical motion analysis systems. J Biomech. 2002;35:537–42. Publisher’sNote 25. Saber-Sheikh K, Bryant EC, Glazzard C, Hamel A, Lee RY. Feasibility of using Springer Nature remains neutral with regard to jurisdictional claims in published inertial sensors to assess human movement. Man Ther. 2010;15:122–5. maps and institutional affiliations. 26. Barbour N, Schmidt G. Inertial sensor technology trends. IEEE Sensors J. 2001;1:332–9. Author details 27. Huddleston J, Alaiti A, Goldvasser D, Scarborough D, Freiberg A, Rubash H, Spinal Cord Injury Center, Heidelberg University Hospital, 69118 Heidelberg, Malchau H, Harris W, Krebs D. Ambulatory measurement of knee motion Germany. Institute for Physiotherapy, University Hospital Jena, 07747 Jena, and physical activity: preliminary evaluation of a smart activity monitor. Germany. Journal of neuroengineering and rehabilitation. 2006;3:21. 28. Motoi K, Tanaka S, Kuwae Y, Yuji T, Higashi Y, Fujimoto T, Yamakoshi K-i. Evaluation Received: 3 January 2018 Accepted: 18 May 2018 of a wearable sensor system monitoring posture changes and activities for use in rehabilitation. Journal of Robotics and Mechatronics. 2007;19:656–66. 29. Grewal GS, Schwenk M, Lee-Eng J, Parvaneh S, Bharara M, Menzies RA, Talal TK, References Armstrong DG, Najafi B. Sensor-based interactive balance training with visual joint 1. Flansbjer U-B, Downham D, Lexell J. Knee muscle strength, gait movement feedback for improving postural stability in diabetics with peripheral performance, and perceived participation after stroke. Arch Phys Med neuropathy: a randomized controlled trial. Gerontology. 2015;61:567–74. Rehabil. 2006;87:974–80. 30. Carpinella I, Cattaneo D, Bonora G, Bowman T, Martina L, Montesano A, 2. Wee J, Lysaght R. Factors affecting measures of activities and participation Ferrarin M. Wearable sensor-based biofeedback training for balance and gait in persons with mobility impairment. Disabil Rehabil. 2009;31:1633–42. in Parkinson disease: a pilot randomized controlled trial. Arch Phys Med 3. Mahlknecht P, Kiechl S, Bloem BR, Willeit J, Scherfler C, Gasperi A, Rungger G, Rehabil. 2017;98:622–30. e623 Poewe W, Seppi K. Prevalence and burden of gait disorders in elderly men and women aged 60–97 years: a population-based study. PLoS One. 2013;8:e69627. 31. Xu J, Bao T, Lee UH, Kinnaird C, Carender W, Huang Y, Sienko KH, Shull PB. Configurable, wearable sensing and vibrotactile feedback system for real- 4. Burnett DR, Campbell-Kyureghyan NH, Topp RV, Quesada PM. Biomechanics time postural balance and gait training: proof-of-concept. Journal of of lower limbs during walking among candidates for total knee arthroplasty neuroengineering and rehabilitation. 2017;14:102. with and without low back pain. Biomed Res Int. 2015;2015:142562. https:// doi.org/10.1155/2015/142562. 32. Afzal MR, Oh M-K, Lee C-H, Park YS, Yoon J. A portable gait asymmetry 5. Michael J, Schlüter-Brust KU, Eysel P. The epidemiology, etiology, diagnosis, rehabilitation system for individuals with stroke using a vibrotactile feedback. and treatment of osteoarthritis of the knee. Dtsch Arztebl Int. 2010;107:152–62. Biomed Res Int. 2015;2015:375638. https://doi.org/10.1155/2015/375638. 6. Mora S, Cook N, Buring JE, Ridker PM, Lee I-M. Physical activity and reduced 33. Shull PB, Jirattigalachote W, Hunt MA, Cutkosky MR, Delp SL. Quantified risk of cardiovascular events. Circulation. 2007;116:2110–8. self and human movement: a review on the clinical impact of wearable Schließmann et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:44 Page 15 of 15 sensing and feedback for gait analysis and intervention. Gait & posture. 57. Cole J, Sedgwick E. The perceptions of force and of movement in a man 2014;40:11–9. without large myelinated sensory afferents below the neck. J Physiol. 1992; 34. Ginis P, Heremans E, Ferrari A, Dockx K, Canning CG, Nieuwboer A. 449:503–15. Prolonged walking with a wearable system providing intelligent auditory 58. Yogev-Seligmann G, Rotem-Galili Y, Mirelman A, Dickstein R, Giladi N, input in people with Parkinson’s disease. Front Neurol. 2017;8:128. Hausdorff JM. How does explicit prioritization alter walking during dual-task performance? Effects of age and sex on gait speed and variability. Phys 35. Dowling AV, Fisher DS, Andriacchi TP. Gait modification via verbal Ther. 2010;90:177–86. instruction and an active feedback system to reduce peak knee adduction 59. Heuninckx S, Wenderoth N, Swinnen SP. Systems neuroplasticity in the moment. J Biomech Eng. 2010;132:071007. aging brain: recruiting additional neural resources for successful motor 36. He Z, Shen Y, Liu T, Yi J, Ferreira JP. A gait retraining feedback system based performance in elderly persons. J Neurosci. 2008;28:91–9. on wearable sensors. In: Advanced Intelligent Mechatronics (AIM), 2017 IEEE 60. Bohannon RW. Comfortable and maximum walking speed of adults aged International Conference on IEEE; 2017. p. 1029–34. 20—79 years: reference values and determinants. Age Ageing. 1997;26:15–9. 37. Vadnerkar A, Figueiredo S, Mayo N, Kearney R. Design and validation of a 61. Brust J, Shafer SQ, Richter RW, Bruun B. Aphasia in acute stroke. Stroke. biofeedback device to improve heel-to-toe gait in seniors. IEEE J Biomed 1976;7:167–74. Health Inf. 2018;22(1):140–6. 62. Pedersen PM, Vinter K, Olsen TS. Aphasia after stroke: type, severity and 38. Donath L, Faude O, Lichtenstein E, Nüesch C, Mündermann A. Validity and prognosis. Cerebrovasc Dis. 2004;17:35–43. reliability of a portable gait analysis system for measuring spatiotemporal 63. Field-Fote EC, Fluet GG, Schafer SD, Schneider EM, Smith R, Downey PA, gait characteristics: comparison to an instrumented treadmill. Journal of Ruhl CD. The spinal cord injury functional ambulation inventory (SCI-FAI). J neuroengineering and rehabilitation. 2016;13:6. Rehabil Med. 2001;33:177–81. 39. Schwesig R, Leuchte S, Fischer D, Ullmann R, Kluttig A. Inertial sensor based 64. Lin J-H, Hsu M-J, Hsu H-W, Wu H-C, Hsieh C-L. Psychometric reference gait data for healthy subjects. Gait & posture. 2011;33:673–8. comparisons of 3 functional ambulation measures for patients with 40. Bellotti F, Kapralos B, Lee K, Moreno-Ger P, Berta R. Assessment in and of stroke. Stroke. 2010;41:2021–5. serious games: an overview. Advances in Human-Computer Interaction. 65. Lord S, Galna B, Rochester L. Moving forward on gait measurement: toward 2013;2013:1. a more refined approach. Mov Disord. 2013;28:1534–43. 41. Moreno-Ger P, Burgos D, Torrente J. Digital games in eLearning environments: current uses and emerging trends. Simulation & Gaming. 2009;40:669–87. 42. Anderson DI, Magill RA, Sekiya H, Ryan G. Support for an explanation of the guidance effect in motor skill learning. J Mot Behav. 2005;37:231–8. 43. Barrios JA, Crossley KM, Davis IS. Gait retraining to reduce the knee adduction moment through real-time visual feedback of dynamic knee alignment. J Biomech. 2010;43:2208–13. 44. van Hedel HJ, Wirz M, Dietz V. Assessing walking ability in subjects with spinal cord injury: validity and reliability of 3 walking tests. Arch Phys Med Rehabil. 2005;86:190–6. 45. Kirshblum SC, Waring W, Biering-Sorensen F, Burns SP, Johansen M, Schmidt-Read M, Donovan W, Graves D, Jha A, Jones L, et al. Reference for the 2011 revision of the international standards for neurological classification of spinal cord injury. J Spinal Cord Med. 2011;34:547–54. 46. Folstein MF, Folstein SE, McHugh PR. “Mini-mental state”: a practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res. 1975;12:189–98. 47. Bohannon RW, Smith MB. Interrater reliability of a modified Ashworth scale of muscle spasticity. Phys Ther. 1987;67:206–7. 48. Mariani B, Rouhani H, Crevoisier X, Aminian K. Quantitative estimation of foot-flat and stance phase of gait using foot-worn inertial sensors. Gait & posture. 2013;37:229–34. 49. Marino RJ, Scivoletto G, Patrick M, Tamburella F, Read MS, Burns AS, Hauck W, Ditunno J Jr. Walking index for spinal cord injury version 2 (WISCI-II) with repeatability of the 10-m walk time: inter- and intrarater reliabilities. Am J Phys Med Rehabil. 2010;89:7–15. 50. Podsiadlo D, Richardson S. The timed “up & go”: a test of basic functional mobility for frail elderly persons. J Am Geriatr Soc. 1991;39:142–8. 51. Demers L, Weiss-Lambrou R, Ska B. The Quebec User Evaluation of Satisfaction with Assist Technol (QUEST 2.0): an overview and recent progress. Technology and Disability. 2002;14:101–5. 52. Graham JE, Karmarkar AM, Ottenbacher KJ. Small sample research designs for evidence-based rehabilitation: issues and methods. Arch Phys Med Rehabil. 2012;93:S111–6. 53. RC Team. R: a language and environment for statistical computing. Vienna; 54. Wickham H. ggplot2: elegant graphics for data analysis: New York: Springer; 55. Saichi K, Yasuda K, Kitaji Y, Kaibuki N, Iwata H. Development and pilot clinical evaluation of a haptic-based perception-empathy biofeedback device for gait rehabilitation. In: Engineering in Medicine and Biology Society (EMBC), 2016 IEEE 38th Annual International Conference of the. IEEE; 2016. p. 6158–61. 56. Vienne A, Barrois RP, Buffat S, Ricard D, Vidal P-P. Inertial sensors to assess gait quality in patients with neurological disorders: a systematic review of technical and analytical challenges. Front Psychol. 2017;8:817.
Journal of NeuroEngineering and Rehabilitation
– Springer Journals
Published: May 29, 2018