Access the full text.
Sign up today, get DeepDyve free for 14 days.
Loading next page...
/lp/springer_journal/a-novel-sensor-based-assessment-of-lower-limb-spasticity-in-children-1Jn2fPxH5r
- Publisher
- Springer Journals
- Copyright
- Copyright © 2018 by The Author(s).
- Subject
- Biomedicine; Neurosciences; Neurology; Rehabilitation Medicine; Biomedical Engineering
- eISSN
- 1743-0003
- DOI
- 10.1186/s12984-018-0388-5
- Publisher site
- See Article on Publisher Site
Abstract
Background: To provide effective interventions for spasticity, accurate and reliable spasticity assessment is essential. For the assessment, the Modified Tardieu Scale (MTS) has been widely used owing to its simplicity and convenience. However, it has poor or moderate accuracy and reliability. Methods: We proposed a novel inertial measurement unit (IMU)-based MTS assessment system to improve the accuracy and reliability of the MTS itself. The proposed system consists of a joint angle calculation algorithm, a function to detect abnormal muscle reaction (a catch and clonus), and a visual biofeedback mechanism. Through spastic knee and ankle joint assessment, the proposed IMU-based MTS assessment system was compared with the conventional MTS assessment system in 28 children with cerebral palsy by two raters. Results: The results showed that the proposed system has good accuracy (root mean square error < 3.2°) and test-retest and inter-rater reliabilities (ICC > 0.8), while the conventional MTS system has poor or moderate reliability. Moreover, we found that the deteriorated reliability of the conventional MTS system comes from its goniometric measurement as well as from irregular passive stretch velocity. Conclusions: The proposed system, which is clinically relevant, can significantly improve the accuracy and reliability of the MTS in lower limbs for children with cerebral palsy. Keywords: Accuracy, Assessment, Cerebral palsy, Inertia measurement unit (IMU), Joint angle, Modified Tardieu scale, Reliability, Spasticity Background also an important tool in determining the effect of inter- Cerebral palsy (CP) is defined as a non-progressive brain ventions, including rehabilitation programs [2], botu- disorder of movement and posture. Most children with linum toxin injections [2, 3], and orthopedic surgeries CP experience spasticity, a motor disorder caused by in- [1]. Moreover, the level of clonus needs to be assessed creased tonic stretch reflexes [1], due to upper motor because clonus can cause instabilities during joint mo- neuron syndrome [2]. CP children have difficulties walk- tions or weight bearings [2, 4, 5]. However, the accuracy ing independently due to abnormal posture and gait, and reliability of the clinical assessments of spasticity are and they have joint deformity and pain in severe cases. quite low owing to the subjectivity of the rater [6]. Of In particular, lower limb spasticity mostly accompanies those, the Modified Ashworth Scale (MAS) is the most clonus, an involuntary, rhythmic, muscular contraction widely used because of its convenience; it is performed and relaxation [2, 3]. with passive stretching by the rater, without any special Spasticity assessment has been used to predict the se- tool. Since the MAS majorly depends on the characteris- verity of CP in activities in daily life (ADL) [1, 4]. It is tics of the resistance felt during the manual passive stretch, the rater highly relies on the subjective feeling, * Correspondence: jhkim@dgist.ac.kr which is sensitive to the passive stretch velocity (PSV) Department of Robotics Engineering, DGIST (Daegu Gyeongbuk Institute of owing to the velocity dependency of spasticity [7, 8], and Science and Technology), 333 Techno Jungang-daero, Daegu 42988, thus has a low reliability. In addition, the MAS has a Republic of Korea 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. Choi et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:45 Page 2 of 16 fundamental limitation in that clonus cannot be assessed evaluation in children with CP. For clinical use, a joint [6]. As a solution to these problems, the Modified Tar- angle calculation algorithm using IMU was developed dieu Scale (MTS) [9], which is described in detail in the without magnetometer. An acceleration mapping method section, was proposed to reduce the rater’s sub- scheme using rotation matrix was included in the algo- jectivity by utilizing a goniometer to consider the vel- rithm for better accuracy. We also developed a method ocity dependence, and this provide a guideline for two for detecting muscle reaction considering clonus (i.e., different PSVs (slow and fast) to assess clonus [10, 11]. duration of clonus). Moreover, for better reliability of Nevertheless, the accuracy of the MTS is still poor and muscle reaction detection, a visual biofeedback based on its reliability has been questioned due to the inaccuracy IMU was developed in the graphic user interface (GUI) of the goniometric measurement and the ambiguous of the proposed system in order to provide regular PSVs (subjective) description of PSV in the guideline, espe- [25]. In addition, we added an IMU attachment guideline cially for the “as fast as possible velocity” [2, 4, 6, 11]. for each target joint to the GUI to maintain the conveni- There were some attempts to improve the spasticity ence of the existing clinical assessments. For validating (and/or clonus) assessment. Several custom devices with the IMU-based MTS assessment using the proposed sys- multiple sensors were developed for more objective as- tem (iMTS), five normal subjects were tested to confirm sessments [12–14]. Since, instead of the MAS or MTS, the accuracy of the joint angle calculation method using less verified parameters which were measured by using IMU, and a clinical trial of 28 patients with CP (18 knee the devices were proposed for spasticity assessment, it and 10 ankle joints) was conducted to evaluate the was difficult to use them immediately in the clinical set- test-retest and inter-rater reliabilities of the iMTS by ting. Few attempts utilized robots or robotic devices to comparing it with the conventional MTS assessment improve the accuracy/reliability of the MAS or MTS and to verify its viability in clinical practice. The origin- [13–16]. However, these were also inadequate to be ap- ality of the proposed iMTS is summarized in Table 1 by plied in the clinical setting, owing to their expensive and comparing the existing IMU-based MTS (and spasticity) complex systems [17]. Moreover, all studies did not assessments. clearly show whether they can be used for clonus assess- ment [12, 14, 15]. Methods Recently, several studies have investigated the MTS by Required functions and characteristics of target joints using the inertial measurement unit (IMU) to improve The MTS assessment procedure [6], except the step the accuracy/reliability problems mentioned above be- measuring range of motion (ROM; called R2 in MTS) is cause an IMU sensor has relatively low cost and is easy illustrated in Fig. 1. Despite fast stretching, the clinician to use [18, 19]. For the MTS assessment, both joint can detect abnormal muscle reaction, including catch angle measurement and muscle reaction (catch or clo- and clonus (#1 and #3 in Fig. 1)[10, 11]. However, it is nus) [2] detection using IMU are essential. Since most difficult to accurately measure the angle of muscle reac- studies used a magnetometer in IMU for an accurate tion (AMR; called R1 in MTS). Hence, iMTS requires 1) joint angle measurement, they were not appropriate in calculation of the joint angle, 2) detection of the location the clinical settings where the heterogeneity of the of occurrence of muscle reaction, and 3) measurement earth’s magnetic fields becomes significant due to ferro- of the duration of clonus (#4 and #5 in Fig. 1). In magnetic and other magnetic materials used in medical addition, it is important 4) to provide regular PSVs dur- devices [19, 20]; to overcome the heterogeneity requires ing fast stretching [25] due to the velocity dependency inconvenient calibration process of the sensor with spe- of spasticity [6]. cial setup [21–23]. A study used a gyroscope without The target joints in this study, namely, the knee and magnetometer, but showed a poor accuracy of joint ankle, mainly exhibit flexion/extension and dorsiflexion/ angle measurement [3]. Moreover, all existing studies on plantarflexion, respectively, in a two-dimensional sagittal IMU-based MTS assessment lacked efforts to provide plane [26]. However, non-sagittal movements are pos- regular PSVs for reliable muscle reaction detection [21]; sible, and some of these movements such as internal/ex- pendulum test [22, 24] and a use of metronome [23] ternal rotation of the knee and inversion/eversion of the were reported to induce raters to control the PSV, but ankle, are not negligible at near full flexion/extension resulted in inaccuracies as well as inconveniences [22]. and dorsiflexion/plantarflexion [26] due to their anatom- In addition, all existing IMU-based MTS assessments ical chracteristics [27]. Considering its clinical import- have a limitation in that clonus assessment was not con- ance, this study focused on the spasticity assessments of sidered [3, 21, 22]. three major muscles: the knee flexor and extensor, and This study proposed a clinically relevant IMU-based ankle plantar-flexor [2]. Each initial posture and passive MTS assessment system to improve the accuracy and re- stretch required for MTS according to the MTS instruc- liability of lower limb (knee and ankle) spasticity tions are summarized in Table 2. Choi et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:45 Page 3 of 16 Table 1 Comparison of MTS (spasticity) assessments using IMU Existing studies Magnetometer Calibration Clonus Regular PSV Accuracy Reliability Target MTS assessment using IMU [21] X X X X O unknown Knee joint (CP n = 20) [23]X X X △ X O Elbow joint (stroke n = 13) [3] O X X X X unknown Ankle joint (CP n =4) Spasticity assessment using IMU [22]X X X △ X O Knee joint (stroke n = 11) [12] O X X X X O Knee / ankle joints (CP n = 28) Proposed system iMTS O O O O O O Knee / ankle joints (CP n = 28) The symbol in the Magnetometer column indicates whether it requires a magnetometer (X) or not (O); the symbol for Calibration indicates whether it requires calibration (X) or not (O); the symbol for Clonus indicates whether it considers clonus (X) or not (O); the symbol for Regular PSV indicate if it is achieved (O), incompletely achieved (△), and not considered (X); the symbol for Accuracy indicates whether the root mean square error of the joint angle is less than 4 deg. (X) or not (O); and the symbol for Reliability denotes whether both test-retest and inter-rater reliabilities are consistent (ICC > 0.8) (X) or not (O) Fig. 1 Schematic diagram of the MTS assessment. Except ROM measurement. SG: spasticity grading; AMR: angle of muscle reaction; AOC: angle of catch; IAOC: initial angle of clonus Choi et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:45 Page 4 of 16 Table 2 Instruction of MTS assessment Joint Knee joint Ankle joint Extensor Flexor Plantar flexor (Quadriceps) (Hamstrings) (Calf muscle) Hip 90 flexion neutral position Knee maximum extension to maximum flexion maximum flexion to maximum extension full extension Ankle N/A maximum plantarflexion to maximum dorsiflexion Note that MTS assessment is conducted in supine position Proposed IMU-based MTS assessment system where θ , θ , and θ denote the segment angles. thigh shank foot Joint angle calculation IMUs were attached to obtain the segment angles in (1), To calculate the joint angle, each segment angle should and the attachment locations and orientations of the be obtained, and then the relative angle should be calcu- IMUs are shown in Table 3 and Fig. 2. The locations lated. According to the definition of each angle shown in were chosen to ensure stable attachment by minimizing Fig. 2, each joint angle can be obtained as follows: skin artifacts [28]. Note that inertial effects, skin deform- ation, and sliding near joints as well as skin deformation due to muscle contraction caused the skin artifacts [29, 30]. θ ¼ θ −θ þ 180 ; knee shank thigh In addition, it should be noted that there are two IMU loca- θ ¼ θ −θ þ 90 ; ankle foot shank tions chosen for shank (Table 3 and Fig. 2); the chosen lo- ð1Þ cation for theankle jointcannot beusedfor thekneejoint Fig. 2 Definition of segment/joint angles and coordinates Choi et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:45 Page 5 of 16 Table 3 Anatomical landmarks for attaching the IMU Knee joint Ankle joint Manipulated Shank: the lateral 1/3 surface of the shank, from the malleolus Foot: the anterior 2nd metatarsal bone segment Held segment Thigh: the lateral 1/3 surface of the thigh, from epicondyle of the Shank: the anterior 1/3 surface of the shank, from the femur tibia because of noticeable skin artifacts (inertial effects and slid- the small thresholds a and ω were experimentally qs_th qs_th ing) caused by significant movement of the shank (manipu- selected as 0.2 m/s and 10 deg./s, respectively. lated segment in the knee joint); further, the location for Assuming that each segment movement occurs only in thekneejoint is also notappropriatefor the anklejoint be- the sagittal plane, the accelerometer can be used as a tilt cause of the skin deformation due to the contraction of the sensor in the quasi-static state (2) [36], and the segment ankle plantar-flexor [29, 31]. The orientations of the IMU angles in the dynamic state can also be obtained by inte- were determined to be as close as possible to the sagittal grating the angular velocity measured from the gyro- plane (global XY plane in Fig. 2). scope as follows [35]: The sensor coordinates of the attached IMUs (x-y-z θ ¼ atan2 a ; a ð3Þ n y n x n axes) are shown in Fig. 2. The method of determining the coordinates was as follows. Since the main movements of Z the target joints (knee flexion/extension and ankle dorsi- θ ¼ ω ðÞ t dt þ θ ð4Þ n z n n latest flexion) appear on the sagittal plane, the normal direction of the plane was set to the z-axes of two attached IMUs where a and a denote the x and the y-axis acceler- x_n y_n [26]. Each x-axis of the IMU coincided with the rotational ation measured by the accelerometer of the IMU at- axis of each additional movement (shank/thigh internal/ tached on the n segment (shank, thigh, and foot), external rotation in knee flexion/extension and foot inver- respectively; ω the z-axis angular velocity measured z_n sion in ankle dorsiflexion) that was not on the sagittal by the gyroscope of the IMU; and θ the latest n_latest plane, as mentioned in the first subsection. Note that the angle obtained by (3) with the accelerometer. However, additional movement of the thigh can be caused by the as mentioned in the first subsection, the significant spasticity pattern of CP [1, 2]. non-sagittal plane movements coexist at near full For clinical use, we used only the accelerometer/gyro- flexion/extension [26]; thus, even if the IMUs were ini- scope of the IMU without the magnetometer to obtain the tially well attached to align the plane (a are near zero), z_n segment angles. The accelerometer is appropriate for they would be outside of the plane. Since the move- quasi-static states [32] but is inaccurate for dynamic states ments occur in the quasi-static state, (3) would result in owing to the additional acceleration caused by its motion a large segment angle error at near full ROM. [33]. Conversely, the gyroscope that can be used in the dy- Hence, when the non-sagittal movements significantly namic state only obtains the relative angle by integrating occurred, we applied a mapping scheme based on a rota- the angular velocity measured [34]; thus, the segment an- tion matrix to obtain the net segment rotation on the sa- gles (absolute angle) depend on the initial value. Moreover, gittal plane (along the Z-axis in the global coordinate) it would have a drift error when used continuously [35]. In only, as illustrated in Fig. 3. Since a (n = 1 or 2) be- z_n this study, we proposed a method to calculate the segment comes non-zero when each segment was outside of the angles by selecting a suitable sensor for each state (quasi-- sagittal plane due to those movements (Fig. 3), we used static or dynamic) that is determined by the measured data the following condition to determine whether the from the IMU [25]. In the quasi-static state, the acceler- scheme was needed: ation measured mainly comes from gravity, and the angular velocity measured was very small. Hence, we defined the a > a ð5Þ z n sp th quasi-static state as if the following conditions were met: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi where a , a small threshold, was heuristically selected qs_th 2 2 2 jg− a þ a þ a j < a qs th 2 x 1 y 1 z 1 as 2 m/s . If (5) was satisfied, the measured IMU accel- ð2Þ erations need to be transformed using the mapping jj ω < ω z 1 qs th scheme. From the definition of rotation matrix based on where a (i = x, y, and z) and ω denote the i-axis linear the z-y-x Euler angles (α, β, and γ)[37], if the rotation of i_1 z_1 acceleration and angular velocity measured by the IMU at- the IMU along the x and y axes (β and γ) is compensated tached in the manipulated segment by the rater. If (2) was by multiplying the rotation matrix, the z-axis of the IMU not met, it was regarded as the dynamic state. Note that coincides with the global Z-axis (Fig. 3) and thus (3) can be Choi et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:45 Page 6 of 16 Fig. 3 Compensation scheme of the x-axis rotation using rotation matrix used to obtain the segment angles on the sagittal plane. calculated by the IMU during the slow stretch can be Here, the mapping of the accelerations using the rotation regarded as the ROM. Compared with goniometric mea- matrix for the compensation was as follows: surements in the MTS, the proposed method can be more convenient to use alone and can increase measure- 2 3 2 3 a^ a x n x n ment accuracy. 4 5 4 5 a^ ¼ R a ; ð6Þ y n y n a^ a z n z n Muscle reaction (catch and clonus) detection 2 3 In addition to the proposed joint angle calculation cαcβ cαsβsγ−sαcγ cαsβcγ þ sαsγ above, the muscle reaction should be detected to obtain 4 5 with R¼ sαcβ sαsβsγ sαsβγ−cαsγ the AMR (Fig. 1). For the MTS, the clinician rapidly ac- −sβ cβsy cβcγ celerates to provide fast PSVs; thus, the joint angular ac- ð7Þ celeration monotonically increases before the muscle reaction. When muscle reaction occurs, the acceleration where script S and U denote the sensor coordinate and is suddenly and greatly decreased due to the reflex the coordinate in which the global coordinate is only ro- torque caused by the muscle reaction (Euler’s 2nd law) tated along the Z-axis, respectively. Note that cα is [39], as displayed in Fig. 5a. Of course, it is possible that cos(α), and sα is sin(α). For (7), we used α = 0 not to the rater adjusts the angular acceleration before the compensate the segment rotation on the sagittal plane, muscle reaction only for PSV control, as shown in Fig. 5b. and β and γ were obtained as follows [33]: However, the acceleration decrease in this unusual case is negligible, compared with the decrease due to the muscle β ¼ atan 2ðÞ a ; a ; z n x n reaction (Fig. 5b). Therefore, we obtained the AMR as the pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð8Þ γ ¼ atan2 a ; a 2 þ a 2 z n x n y n position at which the angular acceleration was minimized between the start and the end of the stretch [39], as shown After the mapping, the segment angles were obtained in Figs. 5 and 6. It allows the clinician to determine the from (3) using the compensated a^ , a^ , and a^ . x n y n z n AOC or IAOC according to the type of the muscle reac- In summary, we proposed a joint angle calculation tion (#3 in Fig. 1). method based on IMU as shown in Fig. 4. It is an effect- Furthermore, it is necessary to measure the duration ive measurement method for knee and ankle motions, of clonus (oscillatory movement) because clonus is di- especially with the non-sagittal plane movements and vided into fatigable (< 10 s) and unfatigable (> 10 s) without a magnetometer. This method can also be used based on the duration (#5 in Fig. 1)[11]. The duration to obtain ROM (R2) in the iMTS. During a slow passive was obtained by measuring the time interval between stretch to measure ROM, the MTS allows the raters to the IAOC to the condition when the angular acceleration stop the stretch when they have reached the subjects’ approaches zero, as displayed in Fig. 6b,and thecondition ROM limit based on the subjects’ haptic feeling [38]. was detected as follows: Hence, the maximum (knee extension and ankle dorsi- σ > 3σ ð9Þ w initial flexion) or the minimum (knee flexion) joint angle Choi et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:45 Page 7 of 16 Fig. 4 IMU-based joint angle calculation algorithm € € where σ denotes the standard deviation of θ (or θ ) spasticity assessment. To achieve the regular PSV, pen- knee ankle in 1-s time window and σ the standard deviation of dulum test [22, 24] and use of metronome [23] were re- initial € € ported, but had many limitations to be used (Table 1). θ (or θ ) during the initial no movement condition. knee ankle The pendulum test based on natural drop due to gravity Figure 6 shows two typical examples of AOC/IAOC requires a fixed initial posture, and causes insufficient and the duration of clonus obtained by the proposed PSV to induce spasticity [22]. The metronome only pro- method. The EMG supports the validity of the detected vides the time duration constantly, and thus it cannot muscle reaction as well as the end of clonus. restrict PSV directly [23]. As such, we developed a visual biofeedback, which has Visual biofeedback recently been reported [25], and included it in a GUI for As mentioned in background chapter, providing regular the proposed iMTS (Table 1), as shown in Fig. 7a.The PSVs is also important to improve the reliability of visual biofeedback was to help the clinician regularly ab Fig. 5 Typical angular acceleration profiles during fast stretching Choi et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:45 Page 8 of 16 ab Fig. 6 Typical knee/ankle joint angles, angular accelerations, and EMG data during fast stretching provide a target PSV, which was selected as a sufficiently performed the stretch within 1 s and felt an abnormal fast velocity for a subject to evoke abnormal muscle re- muscle reaction [10, 11, 40]. actions during passive stretching. Using GUI, it dis- played both the allowable range of PSV (from 90 to Implementation 110% of the target PSV) and the PSV measured by the To implement the proposed IMU-based MTS assess- gyroscope of the IMU (ω + ω ), as a red solid line ment system, two IMUs (shimmer3, the shimmer, z_1 z_2 and blue bar, respectively (Fig. 7a)[26]. To check easily Ireland) consisting of a three-axes accelerometer, gyro- whether the PSV provided was well regulated, we added scope, and magnetometer were used (Table 3). Note that a green indicator that turns on when the maximum ac- the magnetometer of the IMU was not applied in this tual PSV is within the range (Fig. 7a). study. The raw data (acceleration and angular velocity It is noteworthy that determining the target PSV is ω) of the IMUs were collected at a sampling rate of essential for the visual biofeedback. However, it 204.8 Hz, and a second-order Butterworth low-pass filter remains unclear how the PSV for the fast stretch should with 10-Hz cutoff frequency was applied to remove be selected [40]. Hence, in this study, the target PSV noise from the raw data. was determined as the average of the three maximum The raw data collected by the IMU were transmitted to PSVs that were collected by the first clinician’s(rater’s) the PC through a Bluetooth communication. As men- three valid fast stretches wherein the clinician tioned above, the visual biofeedback was implemented as Choi et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:45 Page 9 of 16 Green Target PSV indicator Allowable range of PSV Shank: the lateral 1/3 Foot: the anterior surface of the shank, from 2nd metatarsal bone the malleolus Shank: the anterior Thigh: the lateral 1/3 Sagittal plane 1/3 surface of the surface of the thigh, from shank, from the tibia epicondyle of the femur - Orange indicators Guideline on IMU attachment Fig. 7 Graphic user interface of the proposed system a GUI by LABVIEW (National Instruments, Austin TX, of clonus were not measured in the study. For the USA). The other part of the GUI, which was used to help Mocap, markers were placed to the lower limbs based the clinician attach the IMUs at the initial stage of the on the plug-in-gait model [37, 41], and the stretch mo- iMTS, provided a guideline image on IMU attachment for tions were captured at a sampling rate of 250 Hz. The each target joint and showed whether the attached IMUs IMUs attached and the Mocap were synchronized using were located on the sagittal plane (orange indicators), as a DAQ board (National instruments, Austin TX, USA). displayed in Fig. 7b. The custom program installed in the The reliability study with patients was conducted to PC was developed by MATLAB (MathWorks, Inc., compare the test-retest and inter-rater reliabilities of Natick, MA, USA) to implement the joint angle calcula- conventional MTS (cMTS) and iMTS. Two channels of tion algorithm and muscle reaction detection method for EMG (Trigno wireless EMG, Delsys Inc. USA) were at- measuring the ROM, AMR, and duration of clonus. tached to the target muscles introduced in Table 2 to confirm the existence and the timing of the abnormal Experiments muscle reaction due to spasticity. Experimental setup To evaluate the proposed iMTS, we designed two exper- Participants iments: the accuracy study with healthy subjects and the Five healthy subjects (three males; age 26.0 ± 2.0 years; reliability study with patients. In the accuracy study, only height167.2 ± 6.9 cm; weight 68.2 ± 11.3 kg) without sur- the accuracy of the joint angle obtained by the proposed gery history and pain in the lower limbs participated in IMU-based algorithm was verified by comparing it with the accuracy study. They signed the informed consent the motion capture system (Mocap; Bonita, Vicon, UK) approved by the Daegu Gyeongbuk Institute of Science during the healthy subjects’ active stretch without abnor- and Technology institutional review board (IRB) prior to mal muscle reaction. Note that the AMR and duration the experiment (No. DGIST-160114-HR-005-03). Choi et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:45 Page 10 of 16 As summarized in Tables 4, 28 children with spastic CP participated for the reliability study. Their knee and/ or ankle joints showed spasticity symptoms (catch or clonus) and did not have 1) severe deformities and 2) botulinum toxin within the last 4 months [42]. All guardians of the children gave written informed con- sents approved by the Pusan National University Yang- san Hospital IRB prior to the experiment (No.05– 2015-117). Using the cMTS and iMTS, two experienced clinicians (a medical doctor and a physical therapist) ex- amined the children, along with a volunteer who only performed the goniometric measurements for a blinded test in the cMTS. Please note that the volunteer was a clinical researcher trained to obtain goniometric mea- surements in clinical practice. Protocols In the accuracy study, the subjects were placed in the supine position with markers placed, and two IMUs were attached with straps (Table 3). After several prac- tices, the subjects performed the slow stretch motion as shown in Table 2 and maintained the posture at the end of the ROM for a few seconds to mimic the slow stretch of the cMTS to measure the ROM. Thereafter, the sub- Fig. 8 Protocol of the reliability study. The 1st rater performed the jects performed the fast stretch motion. All subjects preparing session before the test, and the 2nd rater performed the actual test. The 1st rater performed a re-test within 3 days after the acted on their dominant legs and repeated the motions initial test three times. Since it was difficult to capture the passive stretch without marker occlusion owing to the rater [41], we used the voluntary movements for this study. session order, conducting the iMTS sessions after finish- For better simulation of the cMTS, we asked the sub- ing the cMTS sessions (Fig. 8), was required to prevent jects to move the manipulated segment only and to per- the rater from learning to target PSV prior to cMTS form the fast stretch within 1 s [10]. session. The protocol of the reliability study, which included For the cMTS session, the rater then performed the the test-retest and inter-rater reliabilities of the cMTS slow and the fast stretches to measure the ROM (R2) and iMTS, is illustrated in Fig. 8. The first rater (clin- and AMR (R1), respectively, and each stretch was re- ician) attached the IMUs and the EMG sensors to the peated three times [39]. Whenever the rater stopped at subjects and conducted several fast stretches to deter- the end of the ROM or stopped/repositioned the sub- mine the target PSV for the iMTS (see the second sub- jects’ posture at the muscle reaction, the volunteer mea- section of method section on visual biofeedback). Then, sured the joint angle of the posture using a goniometer all raters used the target PSV for visual biofeedback dur- based on the standard measurement method [6]. Note ing fast stretches (Fig. 8). It should be noted that a fixed that we collected the data from the IMUs in this session to investigate how the goniometric measurement affects the reliability of the cMTS. Table 4 Characteristics of study population After the cMTS session, the participants were given an Characteristics Group 1 (n = 18) Group 2 (n = 10) adequate rest period (more than 10 min) to minimize for knee joint for ankle joint the effect of the fixed session order (Fig. 8). Next, the Age (years) 7.5 ± 3.1 5.5 ± 3.5 first rater performed the iMTS session using the visual Weight (kg) 25.1 ± 14.5 15.1 ± 8.3 biofeedback with the target PSV determined earlier. As Height (cm) 119.7 ± 21.1 101.8 ± 23.5 in the cMTS, the iMTS session consisted of three slow Male / Female 11 / 7 6 / 4 and fast stretches (Fig. 8). If the rater failed to provide hemiplegia / bilateral / quadriplegia 5 / 9 / 4 2 / 5 / 3 the target PSV despite the visual biofeedback, additional fast stretch trials were allowed to obtain three valid fast GMFCS I: 3, II: 8, III: 2, I: 1, II: 1, III: 4, IV: 3, V: 2 IV: 0, V: 4 stretch trials with the target PSV [25]. Although the GMFCS: gross motor function classification scale [1] iMTS do not require the goniometric measurement, the Choi et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:45 Page 11 of 16 volunteer additionally did the measurement in this ses- quantitatively evaluated by intraclass correlation coeffi- sion to determine the effect of PSV regulation due to the cients (ICC), which were calculated using the SPSS soft- visual biofeedback of the iMTS. ware version 23 (IBM corporation., USA). Note that Thereafter, the subjects rested for more than ICC > 0.8 indicates good consistency; ICC > 0.6 moderate 15 min, and the second rater conducted the same consistency; and below (ICC < 0.6) poor consistency [43, cMTS and iMTS sessions with the subjects to evalu- 44]. Moreover, to investigate the cause of reliability deteri- ate the inter-rater reliability (Fig. 8). Note that both oration of the cMTS, we additionally calculated the ICCs of raters’ target PSV in the iMTS were identical. To the AMR obtained from 1) the IMU (calculating the joint evaluate the test-retest reliability, the first rater re- angle and detecting the muscle reaction) in the cMTS ses- peated the cMTS and iMTS sessions with the same sions and from 2) the goniometer in the iMTS sessions. subject within 3 days (Fig. 8). The former, cMTS with IMU, was used to show the effect of goniometric measurements and the latter, cMTS with visual biofeedback, to determine the effect of unregulated Data analysis PSVs. In the accuracy study, we compared the joint angle esti- The duration of clonus obtained by the IMU was mated using the proposed algorithm in Fig. 3 with that compared with that measured by the EMG. The measured by Mocap using the Nexus motion analysis RMSE and ICC between the two were obtained for software (Vicon, UK), as displayed in Fig. 9. In the MTS, evaluation. It is noteworthy that the method via EMG the rater measured the ROM after stopping the slow in determining the duration was the same as that (9) stretch and recognized the AMR in the middle of the via IMU [39]. fast stretch. Hence, based on (2), we opted for the In addition, we attempted to confirm the effectiveness quasi-static period in the slow stretch motion to evaluate of the visual biofeedback. It was evaluated by the achiev- the ROM accuracy and the dynamic period in the fast ing rate of the target PSV as follows [25]: stretch motion to evaluate the AMR accuracy (Fig. 9). To evaluate the accuracy quantitatively, the root mean λ ðÞ % ¼ 100 ð10Þ ar square error (RMSE) between two joint angles during total the periods were obtained using the MATLAB. Of the quasi-static periods, the periods that correspond to the where n denotes the required number of stretches to total ankle dorsi-flexor were excluded from the calculation of achieve the target PSV three times. the RMSE (Fig. 9b). We used two-way ANOVA to test for the difference in accuracy of the proposed algorithm Results between the motions (knee extension/flexion and ankle Accuracy study dorsiflexion) that corresponded to the target muscles as The RMSE of the joint angles showing the ROM accur- well as between the outcomes (ROM and AMR). acy (quasi-static periods in slow stretch motions) and According to the MTS assessment procedure shown in AMR accuracy (dynamic periods in fast stretch motions) Fig. 1, the test-retest and inter-rater reliabilities were an- is summarized in Table 5. The RMSEs were less than 4° alyzed using the ROM (R2), AMR (R1), and spasticity in all cases. For all motions, the RMSE for the AMR was angle (SA; difference between R2 and R1) obtained from larger than the RMSE for the ROM, and those RMSEs the cMTS and iMTS sessions. The reliability was were statistically different (p = 0.006), showing the Fig. 9 Representative joint angle comparison between the IMU and Mocap Choi et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:45 Page 12 of 16 difference between the accuracies of ROM and AMR. Achieving rate of the target PSV Compared to knee extension, knee flexion and ankle The calculated achieving rates using (10) are summa- dorsiflexion resulted in larger RMSE. The difference in rized in Table 9. The mean achieving rate was ~ 77%, the RMSEs between the motions was statistically signifi- which indicated that approximately four stretches cant (p = 0.045). have been performed to provide three successful PSVs. Note that the second rater participated only in the test trials (Fig. 8). Reliability study Test-retest and inter-rater reliabilities Discussion The test-retest reliabilities of the cMTS and iMTS are sum- This study was conducted to overcome the limitations of marized in Table 6. The cMTS showed a poor (AMR) to the MTS in terms of its accuracy and reliability by pro- moderate (ROM and SA) consistency for the knee flexor, posing the iMTS. To improve its accuracy, we proposed poor (AMR and SA) to moderate (ROM) consistency for an IMU-based joint angle calculation algorithm as a part the knee extensor, and moderate (AMR and SA) to good of the iMTS. Despite a magnetometer not being used in (ROM) consistency for the ankle plantar-flexor; the iMTS the IMU, the proposed algorithm reduced the joint angle showed good consistency in all cases. measurement error of the cMTS (less than 10° [45]) by Table 6 shows the results of the inter-rater reliability. about 69% (about 3° RMSE in Table 5) on the knee joint. The proposed iMTS showed good consistency in all the The accuracy of the proposed algorithm was comparable muscles (joints), whereas the cMTS showed good to the existing algorithm with magnetometer [21–23], consistency in the ROM of the ankle plantar-flexor only. (about 4° RMSE) and without magnetometer (from 5° to The AMR of the knee flexor and all SA of the cMTS 8° RMSE) [3, 12]. were poor, and the other AMR (knee extensor and ankle The significant difference in the RMSE between the plantar-flexor) and ROM (knee flexor and extensor) of ROM and AMR shows that the AMR error was larger the cMTS showed moderate consistency. than the ROM error. It was because AMR measurement, which was conducted by (4) in the dynamic state, would be vulnerable to drift errors of the gyroscope in the Causes of reliability deterioration in the MTS assessment IMU. However, owing to the short dynamic period re- The comparison on the reliabilities (ICC) of the AMR garding the AMR (< 1 s) in the MTS assessment, the under the cMTS, iMTS, and two modified cMTS by drift error due to the integration in (4) did not occur sig- adding parts of the iMTS is summarized in Table 7.In nificantly (Table 5). The accuracy difference according to all muscles, both the modified cMTS had better the motions would result from the anatomical character- test-retest and inter-rater reliabilities than the cMTS, istics of the joints; the non-sagittal plane movement of and had worse reliabilities than the proposed iMTS. the ankle (inversion) was larger than that of the knee However, it was not clear which modification results in a (internal/external rotation), as mentioned in the first more significant improvement of the reliability. subsection of method section. Since the proposed algo- rithm used a mapping scheme using rotation matrix to compensate for the movements, as shown in (6), (7), and Duration of clonus (8), the RMSEs of the ankle motion were still relatively Table 8 shows the agreement of the duration of clonus small (Table 5). between the iMTS and conventional EMG methods. On In addition to the joint angle calculation algorithm, average, there was a small error (~ 0.07 s), and the high the proposed iMTS contains a muscle reaction detection ICC showed good consistency between both durations. method as well as a visual biofeedback mechanism to Table 5 RMSE between the Mocap and IMU Motion ROM accuracy (deg) AMR accuracy (deg) Mean (SD) RMSE (SD) Mean (SD) RMSE (SD) Knee extension Mocap 165.05 (6.48) 2.24 (1.55) 145.68 (7.31) 2.95 (1.07) IMU 164.16 (7.20) 147.53 (7.47) Knee flexion Mocap 72.92 (5.43) 3.05 (1.84) 104.21 (8.34) 3.97 (2.02) IMU 74.58 (6.91) 105.53 (9.78) Ankle dorsiflexion Mocap 32.92 (5.11) 3.11 (2.91) 5.91 (6.61) 3.86 (1.86) IMU 30.08 (3.03) 7.62 (7.65) SD denotes standard deviation Choi et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:45 Page 13 of 16 Table 6 Test-retest and inter-rater reliabilities Test-retest reliability Inter-rater reliability Test (deg) Re-test (deg) ICC (95% CI) 1st Rater (deg) 2nd Rater (deg) ICC (95% CI) Mean (SD) Mean (SD) Mean (SD) Mean (SD) Knee flexor ROM cMTS 139.19 (10.48) 140.01 (8.59) 0.71 (0.38–0.80) 139.19 (10.48) 139.66 (8.35) 0.65 (0.43–0.77) (R2) iMTS 144.73 (11.61) 142.34 (8.94) 0.84 (0.55–0.92) 144.73 (11.61) 142.34 (8.94) 0.81 (0.65–0.89) AMR cMTS 122.94 (15.74) 120.27 (12.60) 0.50 (0.31–0.70) 122.94 (15.74) 125.88 (15.05) 0.50 (0.17–0.73) (R1) iMTS 126.59 (14.82) 126.06 (14.32) 0.89 (0.80–0.95) 126.59 (14.82) 126.78 (14.75) 0.86 (0.63–0.95) SA cMTS 16.25 (12.38) 19.74 (11.09) 0.63 (0.27–0.78) 16.25 (12.38) 19.74 (11.09) 0.41 (0.48–0.61) (R2-R1) iMTS 18.63 (10.30) 17.76 (10.01) 0.81(0.74–0.91) 18.63 (10.30) 17.47 (10.56) 0.80 (0.53–0.90) Knee extensor ROM cMTS 37.43 (16.16) 46.00 (14.82) 0.73 (0.35–0.79) 37.43 (16.16) 45.01 (19.02) 0.79 (0.50–0.85) (R2) iMTS 47.91 (15.61) 46.94 (9.86) 0.83 (0.73–0.93) 47.91 (15.61) 45.25 (17.28) 0.88 (0.57–0.92) AMR cMTS 39.49 (18.38) 45.76 (14.82) 0.64 (0.32–0.78) 39.49 (18.38) 54.68 (17.82) 0.59 (0.46–0.84) (R1) iMTS 49.97 (16.08) 47.73 (8.92) 0.86 (0.74–0.93) 49.97 (16.08) 58.59 (16.01) 0.84 (0.70–0.92) SA cMTS −4.48 (3.59) −5.15 (4.92) 0.40 (0.24–0.59) − 4.48 (3.59) −9.67 (5.38) 0.50 (0.10–0.73) (R2-R1) iMTS −5.20 (5.43) −4.96 (4.46) 0.81 (0.63–0.91) −5.20 (5.43) −12.13 (9.47) 0.80 (0.54–0.86) Ankle plantar-flexor ROM cMTS 22.77 (7.49) 24.44 (9.50) 0.94 (0.43–0.97) 22.77 (7.49) 21.21 (7.82) 0.94 (0.76–0.91) (R2) iMTS 22.29 (10.57) 21.59 (8.78) 0.92 (0.65–0.89) 22.29 (10.57) 24.78 (8.86) 0.92 (0.66–0.92) AMR cMTS 3.33 (7.57) 4.67 (6.12) 0.71 (0.46–0.78) 3.33 (7.57) 8.24 (9.82) 0.63 (0.38–0.83) (R1) iMTS 8.27 (3.29) 6.40 (4.88) 0.83 (0.55–0.90) 8.27 (3.29) 10.27 (6.92) 0.83 (0.61–0.91) SA cMTS 19.44 (7.72) 19.47 (2.63) 0.63 (0.01–0.45) 19.44 (7.72) 12.97 (9.92) 0.19 (0.04–0.59) (R2-R1) iMTS 14.02 (8.44) 15.19 (9.87) 0.80 (0.68–0.96) 14.02 (8.44) 14.51 (9.46) 0.81 (0.68–0.89) CI denotes the confidence interval improve the reliability by considering the the rater’ssubjectivehapticfeeling to decide theend of the velocity-dependent characteristics of muscle reaction. Our ROM. In fact, the ROM had a lower reliability than the clinical tests showed that the test-retest and inter-rater reli- AMR (Table 6), while the ICCs of the ROM in the iMTS abilities of the proposed iMTS significantly improved com- were higher than the reported ICCs of the conventional pared with those of the cMTS. Moreover, the reliability of ROM measurement (< 0.79) [43]. the iMTS (good consistency in all cases) was better in this From the comparison study, we found that the deteri- studythaninthe existing studiesonimproving theMTS in oration in reliability of the MTS assessment is due to the lower extremities [22]. The SA, which was determined the combined causes of goniometric measurement and by the AMR and ROM, showed slightly lower reliability unregulated PSV. It can be supported that all modified than the AMR (Table 6). It was because the iMTS still used conditions of the cMTS (cMTS with IMU and cMTS with VB) still showed poor to moderate consistency Table 7 Cause of reliability deterioration of AMR in the MTS (Table 7). The fact that there was no dominant cause be- assessment tween the two shows why the iMTS was proposed by in- Test-retest (ICC) Knee flexor Knee extensor Ankle tegrating the joint angle calculation algorithm with plantar-flexor abnormal muscle reaction detection and visual cMTS 0.50 (0.31–0.70) 0.64 (0.32–0.78) 0.71 (0.46–0.78) biofeedback. cMTS with IMU 0.78 (0.51–0.88) 0.74 (0.44–0.90) 0.77 (0.54–0.87) The main outcome of the MTS was the SA, the differ- ence between ROM and AMR, which distinguishes the cMTS with VB 0.71 (0.49–0.79) 0.82 (0.60–0.91) 0.79 (0.67–0.86) neural (dynamic spastic) component from total resistance iMTS 0.89 (0.80–0.95) 0.86 (0.74–0.93) 0.83 (0.55–0.90) (hyper-resistance) during passive stretching [2, 46, 47]. Inter-rater (ICC) Knee flexor Knee extensor Ankle Since it is well-known that the SA is closely related to the plantar-flexor cMTS 0.50 (0.17–0.73) 0.59 (0.46–0.84) 0.63 (0.38–0.83) Table 8 Duration of clonus cMTS with IMU 0.65 (0.49–0.76) 0.78 (0.67–0.84) 0.69 (0.59–0.77) Mean (SD) ICC (95% CI) cMTS with VB 0.76 (0.53–0.81) 0.75 (0.66–0.82) 0.72 (0.61–0.82) iMTS 7.02 (5.01) 0.96 (0.89–0.98) iMTS 0.86 (0.63–0.95) 0.84 (0.70–0.92) 0.83 (0.61–0.91) EMG 7.09 (4.78) All data are presented as ICCs (95% CI); VB denotes visual biofeedback Choi et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:45 Page 14 of 16 Table 9 Achieving rate of the target PSV assessment. For clinical use, the target PSV need to fixed to compare the level of spasticity between subjects. Since 1st rater 2nd rater there is no clear standard to determine PSV for MTS as- Test (%) Re-test (%) Test (%) sessment, especially iMTS, standardization of the target Knee flexor 75 80 78 PSV may be required. The experimental verification of Knee extensor 77 75 83 the accuracy of iMTS was conducted for healthy subjects Ankle plantar-flexor 80 76 77 although there may be a difference in the movements Average (%) 77 between healthy subjects and patients. Even though the proposed joint angle calculation algorithm can compen- therapeutic effect of botulinum toxin type A injection sate some non-sagittal movements (knee internal/exter- (BTX-A) [2, 48], the SA measurement was consistent be- nal rotation and ankle inversion/eversion) that are cause iMTS contributes to 1) successful rehabilitation noticeable even in healthy people due to the anatomical management (e.g., BTX-A without surgery) [49]and 2) characteristics of knee and ankle joints, some bony de- better efficacy and safety of BTX-A through dose adjust- formities of CP children (e.g., foot equinus and hindfoot ment based on reliable SA measurement [50]. Moreover, valgus [52]) can cause non-sagittal movements that the the reliable ROM/AMR measurement of the iMTS is algorithm cannot compensate for. Hence, the low accur- beneficial to physical/occupational therapy and orthotic acy due to those deformities needs to be improved in fu- treatment [51]. ture work. In addition, we need to expand the target In addition to ROM, AMR, and SA, the proposed muscles by including the ankle dorsi-flexor that was ex- iMTS can also evaluate another key part of MTS, clo- cluded owing to its relatively lower clinical importance nus, while clonus was not considered in existing stud- and the upper limb muscles, such as the elbow flexor ies on improving the MTS assessment [3, 21, 22]. For [23] for stroke patients. clonus, the ankle was included to the target joint, and the IMU-based muscle reaction detection method was Conclusions developed to detect the clonus accurately with the In this paper, we proposed a novel sensor-based spasti- catch. We also added the function to measure the city assessment system to improve the accuracy and reli- duration of clonus to distinguish between fatigable ability of a well-known clinical instrument for spasticity, and unfatigable clonus. The accuracy of the measure- namely, MTS. For the IMU-based MTS assessment, with ment was comparable to that of the measurement via consideration to the clinical environment, we developed EMG (Table 8). a magnetometer-free joint angle calculation method, a Compared with existing studies, the present study muscle reaction (catch and clonus) detection function, attempted to develop an improved MTS assessment sys- and a visual biofeedback method to help regulate PSV. tem that is practical enough for clinical use. The pro- The accuracy of the proposed system was validated posed iMTS followed the procedure of the cMTS, and through a comparison with a motion capture system, the sensor attachment location was determined by con- and the reliability of the system was evaluated by con- sidering the anatomical characteristics (Table 3). More- ducting a clinical spasticity assessment of the lower over, the iMTS contains a novel IMU-based joint angle limbs (knee and ankle joints) of 28 children with cere- calculation without magnetometer, considering the clin- bral palsy. With high accuracy of the joint angle calcula- ical setting; the magnetometer for joint angle calculation tion (RMSE < 3 deg), the results of the clinical test requires inconvenient calibration procedure due to local showed that the proposed system can significantly im- distortion of the earth’s magnetic field caused by un- prove test-retest and inter-rater reliabilities of MTS known materials and magnetic objects in outpatients (good consistency; ICC > 0.8) compared to conventional and medical devices, such as ultrasound and transcuta- MTS (poor or moderate consistency; 0.2–0.6 ICC). With neous electrical nerve stimulation [20]. We also devel- the proposed system, it was also noted that the deterior- oped a visual biofeedback to provide regular PSVs for ation in reliability of conventional MTS assessment is clinicians’ convenient use [25]; thus, the high success due to the combination of goniometric measurement rate of providing target PSVs was obtained (Table 9). and unregulated PSV. This study has several limitations. The iMTS assumed that the rater can check whether muscle reaction exists Abbreviations (#1 in Fig. 1) and whether it was clonus (#3 in Fig. 1). AMR: Angle of muscle reaction; AOC: Angle of catch; BTX-A: Botulinum toxin Although this is not a strong assumption, especially type A injection; cMTS: Conventional modified Tardieu scale assessment; CP: Cerebral palsy; GUI: Graphic user interface; IAOC: Initial angle of clonus; when the rater is a well-experienced clinician, the iMTS ICC: Intraclass correlation coefficients; iMTS: IMU-based modified Tardieu needs to determine them autonomously, including the scale assessment; IMU: Inertial measurement unit; IRB: Institutional review spasticity grade [2, 11] for a more objective spasticity boards; MAS: Modified Ashworth scale; Mocap: Motion capture system; Choi et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:45 Page 15 of 16 MTS: Modified Tardieu scale; PSV: Passive stretch velocity; RMSE: Root mean 7. Lance JW. Symposium synopsis. In: Feldman RG, Young RR, Koella WP, square error; ROM: Range of motion; SA: Spasticity angle editors. Spasticity: disordered motor control. Chicago: Year book publishers; 1980. p. 485–94. 8. Patrick E, Ada L. The Tardieu scale differentiates contracture from Acknowledgments spasticity whereas the Ashworth scale is confounded by it. Clin Rehabil. The authors would like to thank all subjects and clinicians volunteered for 2006;20:173–82. the study. 9. Tardieu G. A la recherche d'une technique de mesure de la spasticite. Rev Neurol. 1954;91:143–4. Funding 10. Scholtes VA, Becher JG, Beelen A, Lankhorst GJ. Clinical assessment of This research was supported by the R&D grant (No.2016002 and 2017002) on spasticity in children with cerebral palsy: a critical review of available rehabilitation by Korea National Rehabilitation Center Research Institute, instruments. Dev Med Child Neurol. 2006;48:64–73. Ministry of Health & Welfare. This work was supported by the DGIST R&D 11. Haugh A, Pandyan A, Johnson G. A systematic review of the Tardieu scale Program of the Ministry of Science and ICT (18-BD-0401). for the measurement of spasticity. Disabil Rehabil. 2006;28:899–907. 12. Bar-On L, Aertbeliën E, Wambacq H, Severijns D, Lambrecht K, Dan B, Availability of data and materials Huenaerts C, Bruyninckx H, Janssens L, Van Gestel L. A clinical measurement Data and materials can be made available upon request to the authors. to quantify spasticity in children with cerebral palsy by integration of multidimensional signals. Gait Posture. 2013;38:141–7. Authors’ contributions 13. Lunenburger L, Colombo G, Riener R, Dietz V. Clinical assessments JK supervised the study. SC, YBS, and JK conceptualized and designed the performed during robotic rehabilitation by the gait training robot Lokomat. study. JK acquired the funding and provided the resources for the study. SC In: Rehabilitation Robotics, 2005 ICORR 2005 9th International Conference and JK developed the proposed assessment system. SC implemented the on: IEEE; 2005. p. 345–8. https://doi.org/10.1109/ICORR.2005.1501116. proposed system. SC and JH designed the experiments. YBS and SK recruited 14. Peng Q, Park H-S, Shah P, Wilson N, Ren Y, Wu Y-N, Liu J, Gaebler-Spira DJ, subjects and prepared IRB for the experiments. SC, YBS, and SK acquired the Zhang L-Q. Quantitative evaluations of ankle spasticity and stiffness in data. SC and JK processed and analyzed the data from the experiments. All neurological disorders using manual spasticity evaluator. J Rehabil Res Dev. authors interpreted results from the data. SC drafted the original manuscript. 2011;48:473. JK finalized the manuscript. All authors read and revised the manuscript, and 15. Mirbagheri MM, Alibiglou L, Thajchayapong M, Rymer WZ. Muscle and approved the final manuscript for publication. reflex changes with varying joint angle in hemiparetic stroke. J Neuroeng Rehabil. 2008;5:6. Ethics approval and consent to participate 16. Alibiglou L, Rymer WZ, Harvey RL, Mirbagheri MM. The relation between The healthy subjects signed an informed consent approved by the Daegu Ashworth scores and neuromechanical measurements of spasticity Gyeongbuk Institute of Science and Technology IRB prior to the experiment following stroke. J Neuroeng Rehabil. 2008;5:18. (No. DGIST-160114-HR-005-03). All guardians of the CP children gave written 17. Maggioni S, Melendez-Calderon A, van Asseldonk E, Klamroth-Marganska V, informed consents approved by the Pusan National University Yangsan Hos- Lünenburger L, Riener R, van der Kooij H. Robot-aided assessment of lower pital IRB prior to the experiment (No. 05–2015-117). extremity functions: a review. J Neuroeng Rehabil. 2016;13:72. 18. Mayagoitia RE, Nene AV, Veltink PH. Accelerometer and rate gyroscope Competing interests measurement of kinematics: an inexpensive alternative to optical motion The authors declare that they have no competing interests. analysis systems. J Biomech. 2002;35:537–42. 19. Roetenberg D, Luinge HJ, Baten CT, Veltink PH. Compensation of magnetic disturbances improves inertial and magnetic sensing of human body Publisher’sNote segment orientation. IEEE Trans Neural Syst Rehabil Eng. 2005;13:395–405. Springer Nature remains neutral with regard to jurisdictional claims in 20. Roetenberg D, Baten CT, Veltink PH. Estimating body segment orientation published maps and institutional affiliations. by applying inertial and magnetic sensing near ferromagnetic materials. IEEE Trans Neural Syst Rehabil Eng. 2007;15:469–71. Author details 21. van den Noort JC, Scholtes VA, Harlaar J. Evaluation of clinical spasticity Department of Robotics Engineering, DGIST (Daegu Gyeongbuk Institute of assessment in cerebral palsy using inertial sensors. Gait Posture. 2009;30:138–43. Science and Technology), 333 Techno Jungang-daero, Daegu 42988, 22. Sterpi I, Caroli A, Meazza E, Maggioni G, Pistarini C, Colombo R. Lower limb Republic of Korea. Department of Rehabilitation Medicine, Pusan National spasticity assessment using an inertial sensor: a reliability study. Physiol University School of Medicine and Biomedical Research Institute, Pusan Meas. 2013;34:1423. National University Hospital, 179 Gudeok-ro, Busan 49241, Republic of Korea. 23. Paulis WD, Horemans HL, Brouwer BS, Stam HJ. Excellent test–retest and Department of Rehabilitation Medicine, Pusan National University Yangsan inter-rater reliability for Tardieu scale measurements with inertial sensors in Hospital, 20 Geumo-ro, Yangsan 50612, Republic of Korea. elbow flexors of stroke patients. Gait Posture. 2011;33:185–9. Received: 11 September 2017 Accepted: 14 May 2018 24. Szopa A, Domagalska–Szopa M, Kidoń Z, Syczewska M. Quadriceps femoris spasticity in children with cerebral palsy: measurement with the pendulum test and relationship with gait abnormalities. J Neuroeng Rehabi. 2014;11:166. References 25. Choi S, Kim J. Improving modified tardieu scale assessment using inertial 1. Krigger KW. Cerebral palsy: an overview. Am Fam Physician. 2006;73(1): measurement unit with visual biofeedback. In: Engineering in Medicine and 91–100. Biology Society (EMBC), 2016 IEEE 38th Annual International Conference of 2. Boyd RN, Graham HK. Objective measurement of clinical findings in the use the: IEEE; 2016. p. 4703–6. https://doi.org/10.1109/EMBC.2016.7591777. of botulinum toxin type a for the management of children with cerebral 26. Wu G, Siegler S, Allard P, Kirtley C, Leardini A, Rosenbaum D, Whittle M, D palsy. Eur J Neurol. 1999;6:S23–35. D’Lima D, Cristofolini L, Witte H. ISB recommendation on definitions 3. Lin Y-C, Lin I-L, Chou T-FA, Lee H-M. Quantitative evaluation for spasticity of of joint coordinate system of various joints for the reporting of human joint calf muscle after botulinum toxin injection in patients with cerebral palsy: a motion—part I: ankle, hip, and spine. J Biomech. 2002;35:543–8. pilot study. J Neuroeng Rehabil. 2016;13:25. 27. Neumann DA. Kinesiology of the musculoskeletal system: foundations for 4. Barnes MP, Johnson GR. Upper motor neurone syndrome and spasticity: physical rehabilitation. St. Louis: Mosby; 2002. clinical management and neurophysiology. 2nd ed: Cambridge, Cambridge University Press; 2008. 28. Cappozzo A, Catani F, Della Croce U, Leardini A. Position and orientation in 5. Hidler JM, Rymer WZ. A simulation study of reflex instability in spasticity: space of bones during movement: anatomical frame definition and origins of clonus. IEEE Trans Rehabil Eng. 1999;7:327–40. determination. Clin Biomech. 1995;10:171–8. 6. Mehrholz J, Wagner K, Meißner D, Grundmann K, Zange C, Koch R, Pohl M. 29. Leardini A, Chiari L, Della Croce U, Cappozzo A. Human movement analysis Reliability of the modified Tardieu scale and the modified Ashworth scale in adult using stereophotogrammetry: Part 3. Soft tissue artifact assessment and patients with severe brain injury: a comparison study. Clin Rehabil. 2005;19:751–9. compensation. Gait Posture. 2005;21:212–25. Choi et al. Journal of NeuroEngineering and Rehabilitation (2018) 15:45 Page 16 of 16 30. Karlsson D, Tranberg R. On skin movement artefact-resonant frequencies of skin markers attached to the leg. Hum Mov Sci. 1999;18:627–35. 31. Alexander EJ, Andriacchi TP. Correcting for deformation in skin-based marker systems. J Biomech. 2001;34:355–61. 32. Luinge HJ, Veltink PH. Inclination measurement of human movement using a 3-D accelerometer with autocalibration. IEEE Trans Neural Syst Rehabil Eng. 2004;12:112–21. 33. Tuck K. Tilt sensing using linear accelerometers. In: Freescale semiconductor application note AN3107; 2007. 34. Cooper G, Sheret I, McMillian L, Siliverdis K, Sha N, Hodgins D, Kenney L, Howard D. Inertial sensor-based knee flexion/extension angle estimation. J Biomech. 2009;42:2678–85. 35. Luinge HJ, Veltink PH. Measuring orientation of human body segments using miniature gyroscopes and accelerometers. Med Biol Eng Comput. 2005;43:273–82. 36. Song M, Kim J. Simple ambulatory gait monitoring system using a single IMU for various daily-life gait activities. In: Biomedical and Health Informatics (BHI), 2016 IEEE-EMBS International Conference on: IEEE; 2016. p. 430–3. https://doi.org/10.1109/BHI.2016.7455926. 37. Winter DA. Biomechanics and motor control of human movement. 4th ed, Hoboken, NJ, USA: John Wiley and Sons; 2009. 38. Stuberg WA, Fuchs RH, Miedaner JA. Reliability of goniometric measurements of children with cerebral palsy. Dev Med Child Neurol. 1988; 30:657–66. 39. van den Noort JC, Scholtes VA, Becher JG, Harlaar J. Evaluation of the catch in spasticity assessment in children with cerebral palsy. Arch Phys Med Rehabil. 2010;91:615–23. 40. WU YN, Ren Y, Goldsmith A, Gaebler D, Liu SQ, ZHANG LQ. Characterization of spasticity in cerebral palsy: dependence of catch angle on velocity. Dev Med Child Neurol. 2010;52:563–9. 41. Moeslund TB, Hilton A, Krüger V. A survey of advances in vision-based human motion capture and analysis. Comput Vis Image Underst. 2006; 104:90–126. 42. Grazko MA, Polo KB, Jabbari B. Botulinum toxin a for spasticity, muscle spasms, and rigidity. Neurology. 1995;45:712–7. 43. Ben-Shabat E, Palit M, Fini NA, Brooks CT, Winter A, Holland AE. Intra-and interrater reliability of the modified Tardieu scale for the assessment of lower limb spasticity in adults with neurologic injuries. Arch Phys Med Rehabil. 2013;94:2494–501. 44. Cicero MX, Riera A, Northrup V, Auerbach M, Pearson K, Baum CR. Design, validity, and reliability of a pediatric resident JumpSTART disaster triage scoring instrument. Acad Pediatr. 2013;13:48–54. 45. Gracies J-M, Burke K, Clegg NJ, Browne R, Rushing C, Fehlings D, Matthews D, Tilton A, Delgado MR. Reliability of the Tardieu scale for assessing spasticity in children with cerebral palsy. Arch Phys Med Rehabil. 2010;91:421–8. 46. Ward AB, Aguilar M, Beyl ZD, Gedin S, Kanovsky P, Molteni F, Wissel J, Yakovleff A. Use of botulinum toxin type a in management of adult spasticity-a European consensus statement. J Rehabil Med. 2003;35:98–9. 47. van den Noort JC, Bar-On L, Aertbeliën E, Bonikowski M, Brændvik SM, Broström EW, Buizer AI, Burridge JH, Campenhout A, Dan B. European consensus on the concepts and measurement of the pathophysiological neuromuscular responses to passive muscle stretch. Eur J Neurol. 2017;24:981. 48. Love S, Novak I, Kentish M, Desloovere K, Heinen F, Molenaers G, O’flaherty S, Graham H. Botulinum toxin assessment, intervention and after-care for lower limb spasticity in children with cerebral palsy: international consensus statement. Eur J Neurol. 2010;17:9–37. 49. Deon LL, Gaebler-Spira D. Assessment and treatment of movement disorders in children with cerebral palsy. Orthop Clin N Am. 2010;41: 507–17. 50. Strobl W, Theologis T, Brunner R, Kocer S, Viehweger E, Pascual-Pascual I, Placzek R. Best clinical practice in botulinum toxin treatment for children with cerebral palsy. Toxins. 2015;7:1629–48. 51. Umphred DA, Lazaro RT, Roller M, Burton G. Neurological Rehabilitation. 6th ed, St. Louis: Elsevier/Mosby; 2013. 52. Koman LA, Smith BP, Shilt JS. Cerebral palsy. Lancet. 363:1619–31.
Journal
Journal of NeuroEngineering and Rehabilitation – Springer Journals
Published: Jun 4, 2018