- Email: [email protected]
Implementation of a markerless motion analysis method to quantify hyperkinesis in males with fragile X syndrome
Gait & Posture 39 (2014) 827–830
Contents lists available at ScienceDirect
Gait & Posture journal homepage: www.elsevier.com/locate/gaitpost
Short Communication
Implementation of a markerless motion analysis method to quantify hyperkinesis in males with fragile X syndrome Joan A. O’Keefe a, Alejandro A. Espinoza Orı´as b, Hassan Khan b, Deborah A. Hall c, Elizabeth Berry-Kravis c,d,e, Markus A. Wimmer a,b,* a
Department of Anatomy and Cell Biology, Rush University Medical Center, Chicago, IL 60612, United States Motion Analysis Laboratory, Department of Orthopedic Surgery, Rush University Medical Center, Chicago, IL 60612, United States Department of Neurological Sciences, Rush University Medical Center, Chicago, IL 60612, United States d Department of Pediatrics, Rush University Medical Center, Chicago, IL 60612, United States e Department of Biochemistry, Rush University Medical Center, Chicago, IL 60612, United States b c
A R T I C L E I N F O
A B S T R A C T
Article history: Received 21 November 2012 Received in revised form 14 October 2013 Accepted 19 October 2013
Hyperactive behavior – and implicitly, motion – in Fragile X syndrome (FXS) has been historically described using behavioral rating scales. While rating scales are the current standard outcome measures used in clinical research, they have limitations including their qualitative nature and subjectivity. The advent of new motion capture technologies has provided the opportunity to develop quantitative methods for measuring hyperactive motion. The hypotheses for this study were that a novel markerless motion analysis method (1) can quantitatively measure kinematic parameters, (2) can differentiate the level of hyperkinesis between control and FXS populations, and (3) will correlate with blind-reviewer synchronous video-capture methods and behavioral rating scale scores. Twenty young males (7-control, 13-FXS; ages 9–32) were studied using a standardized protocol in a markerless motion analysis suite. Behavioral scale questionnaires were filled out by parents and those scores were correlated with motion parameters (frequency and total traveled distance) of body segments during 30 s of story listening while standing in the observation space. The markerless system was able to track subjects and the two populations displayed significantly different quantities of motion, with larger amounts of motion in the FXS group (p < 0.05). Pearson’s correlation coefficients between frequency counts obtained from the markerless system and rater-based video capture were between 0.969 and 0.996 (p < 0.001). Significant correlations between rating scale scores and motion parameters ranged from 0.462 r 0.568 (p 0.040). These results suggest feasibility and validity of a markerless system as a non-invasive method able to quantify motion in individuals with hyperkinesis. ß 2013 Elsevier B.V. All rights reserved.
Keywords: Hyperkinesis Hyperactivity Fragile X syndrome Markerless motion analysis Behavioral rating scales
1. Introduction Fragile X syndrome (FXS) is the most common inherited cause of intellectual disability and the most common known genetic cause of autism [1]. It is associated with increased motor activity (hyperkinesis) that includes hyperactivity with increased body and limb movement and motor stereotypies (repetitive, non-goal directed movement patterns). Hyperactivity has been quantified in rodent models using open field tests to measure locomotor activity of the animal in a defined space [2]. An equivalent quantitative method does not exist for humans. Current outcome tools used
* Corresponding author at: Motion Analysis Laboratory, Rush University Medical Center, 1611 W Harrison Street, Suite 201 Orthopaedic Building, Chicago, IL 60612, United States. Tel.: +1 312 942 2789; fax: +1 312 942 2040. E-mail address: [email protected] (M.A. Wimmer). 0966-6362/$ – see front matter ß 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.gaitpost.2013.10.017
when investigating pharmaceutical (or other therapy) effects on hyperkinesis typically use parent/guardian rating scales [3,4]. These scales are the present gold standard for use in clinical trials research [5] and have validity and reliability [6,7]. However, they are not ideal because they are qualitative, inherently subjective, unable to detect small increments of behavioral changes, and prone to floor and/or ceiling effects. Traditional state of the art marker based motion analysis approaches to analyze hyperkinetic behaviors have not been reported in the FXS or autism literature, but they would likely have considerable limitations in this developmentally disabled population. Marker placement requires an expert spending a considerable amount of time with marker application and a subject who understands, does not object to removing clothing, and is cooperative with the instructions for motion analysis. A subject population with FXS and/or autism is expected to be non cooperative with the tedious marker application process due to behavioral issues including tactile
828
J.A. O’Keefe et al. / Gait & Posture 39 (2014) 827–830
defensiveness, anxiety, and lack of tolerability and understanding of the procedure. The above limitations in quantifying hyperkinesis and stereotypies, in combination with the critical need to quantify these core behavioral features of FXS and autism for efficacy research provided the impetus for the present study. Motion analysis is used in several fields of medical practice, including orthopedics, sports medicine, rehabilitation, and neurology. Accurate description of body kinematics is typically achieved with the use of reflective markers placed on anatomical landmarks to track limb, head and trunk position. Only recently have technological advances in image-recognition algorithms allowed development of systems robust enough to be used in clinical, sports and occupational biomechanics [8–11]. The objective of this study was to determine the feasibility of using markerless motion technology to quantify motor activity (specifically body mass center and limb segment movement) in young males with and without FXS. Our hypotheses are that a markerless motion analysis system will provide quantitative measures of hyperkinesis observed clinically in subjects with FXS, differentiate levels of motor activity between controls and those with FXS, and motion parameters of hyperkinesis will correlate with both videocaptured and clinical hyperactivity rating scale scores. Ultimately, we anticipate that this quantitative system will serve as a model for markerless motion measurement in hyperkinetic disorders. 2. Materials and methods The markerless motion capture system (Biostage, Organic Motion, New York, NY) acquired motion with fourteen video cameras mounted in a self-contained observation space bound by three reflective-canvas walls measuring with a footprint of 3.5 m 4.9 m and a wall height of 2.5 m (Fig. 1a). This system has been reported to have high correlation factors with a traditional marker-based system (VICONTM) for kinematic gait variables [12]. The 3D motion data was analyzed using The MotionMonitor1 inverse-dynamics analysis suite (Innsport, Chicago, IL). The following movement parameters are reported: (1) traveled distance of the body center-of-mass (COM), (2) motion frequency of the forearm and foot, and (3) traveled distance of the forearm and foot. Per convention, the right side was selected as the index body segment. Twenty male subjects including seven controls (age 9–22, mean 14.2 5.14 years) and thirteen FXS patients (age 10–32, mean 15.3 7.27 years) were recruited. The motion data was recorded during a standardized protocol. A variety of toys, manipulatives, and books spanning a broad age range of interest were offered to subjects
before actual motion recordings took place. Subjects were allowed to choose a favorite item before stepping into the Biostage area. Each subject had to go through a series of static standing postures (e.g. a Tpose during which the arms had to be held parallel to the ground for 2–5 s), to register the subject with the software and calibrate the system. Holding their favorite item, the subjects were then asked to listen to an audio-excerpt from Dr. Seuss’ ‘‘The Cat in the Hat’’ which has wide universal appeal to many age ranges. The audioclip was chosen because we empirically found it to hold the interest of all age ranges tested in this pilot study, including children with FXS who have intellectual difficulties and children and adult controls. The subjects were not instructed to stand still but were simply asked to listen to the story. The markerless system recorded motion data during three consecutive thirty-second intervals of story listening. Hyperactivity rating scales, including the Aberrant Behavior Checklist (ABC)-C – Hyperactivity Subscale score [3] and the Swanson, Nolan and Pelham IV (SNAP IV) – Hyperactivity/ Impulsivity subscale score [4] were filled out by parents on the day of testing and were correlated with the motion parameters in order to determine validity of the motion parameter outcomes. Further validity and reliability testing was done by having simultaneous, synchronized videotaped analysis of one control and one FXS subject’s motion during the markerless testing protocol repeated over three sessions. Concurrent criterion validity was tested by having 3 reviewers who were blind to the subject’s diagnosis perform off-line scoring of the frequency of eight motion parameters obtained from the video-capture method and then determining the correlations between these scores and those from the markerless system viewed on the TMM software interface. These included number of steps taken with the feet and arm movement in both upward and downward directions. Test–retest reliability of motion parameters was performed by examining the repeatability of the traveled distance measures obtained in the markerless system across the three sessions. The study was approved by the Rush University Institutional Review Board. All statistical analyses were performed using SPSS v 19.0 (SPSS Inc. Chicago, IL). Motion data from subjects were log-normalized before analysis. Independent student t-tests were then used for group comparisons. Correlations between the behavioral scales and the markerless kinematics parameters were obtained using two-tailed Spearman’s correlations. Two-tailed Pearson’s correlations coefficients were used to compare the kinetic data from the markerless system with the observer based counts from the video. Repeatability of the markerless motion parameters was determined using intraclass correlation coefficients. Statistical significance was established at a 0.05.
Fig. 1. (a) Depiction of the test area (Biostage). Three-dimensional view of the skeleton and center of body mass trajectory of representative, (b) FXS and (c) control subjects, respectively.
J.A. O’Keefe et al. / Gait & Posture 39 (2014) 827–830
829
Fig. 2. Traveled distances (in meters) of the (a) right forearm and (b) right foot relative to body mass center during 30 s of story listening. Note the differences in median and range between the control and FXS groups. The data are plotted on a logarithmic scale.
3. Results FXS subjects showed a large amount of movement in the Biostage. Fig. 1b and c show skeleton views of representative subjects from both experimental groups as viewed on the Motion Monitor’s interface. Arm and foot travel distances, were significantly greater in the FXS group compared to controls (p = 0.038 and 0.011, respectively, Fig. 2). The distance traveled by the body COM was greater in the FXS (3.3 4.0 m) versus control (1.1 0.5 m) groups. Due to the variable activity pattern within the FXS subjects, this difference did not reach significance (p = 0.177). All motion parameters were significantly associated with the ABC-C Hyperactivity subscale score (r = 0.481–0.568; p 0.032) and all except forearm frequency were significantly associated with the SNAP IVHyperactivity/Impulsivity Subscale score (r = 0.462–0.558; p 0.040; Table 1). Stronger correlations were obtained from travel distance parameters than motion frequency values. Motion parameters from the markerless system were significantly correlated with those counts obtained from the three blinded reviewers of simultaneous video-capture. The Pearson correlation coefficients ranged from 0.969 to 0.996 (p < 0.001). Repeated analysis of body segment movements generated similar total traveled distance values in all sessions. The 95% confidence interval of the intraclass correlation coefficients ranged from 0.919 to 0.986 (p < 0.001), thus providing further confidence in the repeatability of the method. 4. Discussion Kinematic assessment of body segments could be useful for future interventional studies attempting to quantify changes in hyperkinesis in patients with hyperkinetic disorders. This study is the first to attempt measurement of kinematic parameters obtained from a markerless motion capture system and correlate these measurements with scores on video-capture methodology
and behavioral rating scales in persons with FXS, a condition associated with hyperkinesis. The results obtained are encouraging because the method was able to discriminate FXS and control groups based on relative body segment travel. This was found to be a more sensitive hyperkinesis marker than ‘COM travel’, which did not reach significance, most likely due to low subject number and the variability of motion parameters within the FXS group. Clinically, FXS subjects move their arms frequently and/or repetitively. These are movements our control subjects never had. However, while standing, some control subjects swayed their upper body from one leg to the other, thus increasing the COM movement. This may have decreased the difference in COM travel distance between control and FXS groups to a non-significant value, despite the fact that many of the FXS subjects walked around excessively in the Biostage. The high correlations obtained between videotape counts from blinded raters and those from the markerless system demonstrate that we have excellent agreement between these two methodologies, suggesting validity of this new technology to quantify hyperkinesis. In addition, the traveled distance of body segments (a kinematic variable not computable from standard video analysis) showed high correlations with behavioral ratings of hyperkinesis, further strengthening the validity of this method in our subject population. While we did not provide proof of the accuracy of the measured distances, we demonstrated that these measures are reproducible and can be used to distinguish control from Fragile X subjects. Further, as mentioned already under Section 2, an earlier study found good agreement in kinematic measures between a marker-based and the markerless system, reporting correlation factors from 0.89 to 1.0 for the sagittal plane motion of the ankle, knee and hip joint [12]. The markerless motion system used in this study has important features and perhaps advantages over a traditional marker-based approach that made it feasible to measure hyperkinesis in persons with FXS. The system does not require the attachment of any leads
Table 1 Spearman correlation coefficients relating behavioral rating scores to motion parameters. Each table cell shows on the top row the correlation coefficient and below the significance value for a two-tailed correlation analysis. Scale/correlation coefficient
Forearm frequency
Foot frequency
Foot travel
Forearm travel
ABC-C (Hyperactivity)
0.533* 0.015
0.481* 0.032
0.568** 0.009
0.537* 0.015
SNAP-IV (Hyperactivity/Impulsivity)
0.330 0.155
0.462* 0.040
0.558* 0.011
0.508* 0.022
* **
Denotes significance at the 0.05 level (two-tailed). Denotes significance at the 0.01 level.
830
J.A. O’Keefe et al. / Gait & Posture 39 (2014) 827–830
or markers and the Biostage has minimal physical constraints, except for the requirement of the individual to stay in the testing area. These qualities increased the subject’s ability to participate in the testing protocol and minimized refusals to cooperate, thus allowing us to obtain quantitative results. Our experimental design was modeled after the open field hyperactivity measure utilized in behavioral studies of the Fragile X knockout mouse model of FXS [2,13,14]. In this test animals are placed in an enclosed area (the ‘‘open field’’) and their motion trajectory is tracked. The markerless motion capture system presented here provides similar behavioral analyses in humans, potentially facilitating translational research. It offers the additional advantage of quantifying the motion of specific body segments, and therefore is more relevant for tracking the manifestations of hyperkinesis in humans. While previous researchers have used video-capture methods synchronized to forceplate data in combination with an optical marker based approach to determine COM trajectory and velocity [15], this methodology requires significant cooperation and understanding on the part of the subject which would have been impossible in our developmentally disabled population. In summary, our feasibility study suggests that motion parameters obtained from the markerless system can quantify increased movement in subjects with FXS relative to controls. Preliminary findings demonstrate that hyperkinesis measures derived from this system are correlated with video-capture based methods and scores on standard behavioral checklists in both controls and individuals with FXS. These data require verification in larger samples with stringent reproducibility and clinical validity testing. Markerless motion capture of hyperkinesis may have important ramifications for removing current barriers in quantifying a major disruptive, phenotypic characteristic in FXS, autism, and other neurodevelopmental disorders. This technology has the potential to be used as a quantitative outcome measure for clinical research trials investigating methods to decrease hyperkinesis in these patient populations. Acknowledgements This study was supported by a pilot grant from the FRAXA Research Foundation (PI: Berry-Kravis). The authors also thank Mr. Robert Trombley from the Motion Analysis Lab at Rush University for his expert assistance with the motion analysis routines. Further,
we would like to thank Renee Kawecki, Chris Ferrigno, Gary Farkas, and Samir Chabra for their assistance with video analysis. Conflict of interest statement None of the authors have financial or other conflicts of interest in regards to this research. References [1] Wang LW, Berry-Kravis E, Hagerman RJ. Fragile X: leading the way for targeted treatments in autism. Neurotherapeutics 2010;7:264–74. [2] Min WW, Yuskaitis CJ, Yan Q, Sikorski C, Chen S, Jope RS, et al. Elevated glycogen synthase kinase-3 activity in Fragile X mice: key metabolic regulator with evidence for treatment potential. Neuropharmacology 2009;56:463–72. [3] Aman MG. Aberrant behavior checklist – community. East Aurora, New York: Slosson Educational Publications; 1994. [4] Swanson JM. School-based assessments and interventions for ADD students. Irvine, CA: KC Publishing; 1992. [5] Berry-Kravis E, Krause SE, Block SS, Guter S, Wuu J, Leurgans S, et al. Effect of CX516, an AMPA-modulating compound, on cognition and behavior in fragile X syndrome: a controlled trial. J Child Adolesc Psychopharmacol 2006;16:525–40. [6] Bihm EM, Poindexter AR. Cross-validation of the factor structure of the aberrant behavior checklist for persons with mental retardation. Am J Ment Retard 1991;96:209–11. [7] Miller ML, Fee VE, Jones CJ. Psychometric properties of ADHD rating scales among children with mental retardation. Res Dev Disabil 2004;25:477–92. [8] Corazza S, Mu¨ndermann L, Chaudhari AM, Demattio T, Cobelli C, Andriacchi TP. A markerless motion capture system to study musculoskeletal biomechanics: visual hull and simulated annealing approach. Ann Biomed Eng 2006;34:1019–29. [9] Moeslund TB, Hilton A, Kru¨ger V. A survey of advances in vision-based human motion capture and analysis. Comput Vision Image Understanding 2006;104:90–126. [10] Ray SJ, Teizer J. Real-time construction worker posture analysis for ergonomics training. Adv Eng Informatics 2012;26:439–55. [11] Urtasun R, Fleet DJ, Fua P. 3D people tracking with gaussian process dynamical models, IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Conf Proc 2006;1:238–45. [12] Oberla¨nder KD, Bru¨ggemann GP. Validation of a real-time markerless tracking system for clinical gait analysis ad-hoc results. In: 35th Annual Meeting of the American Society of Biomechanics. Conference Proceedings; 2011. Paper #304 (http://www.asbweb.org/conferences/2011/pdf/304.pdf; accessed on 13.10.13). [13] Yan QJ, Rammal M, Tranfaglia M, Bauchwitz RP. Suppression of two major Fragile X syndrome mouse model phenotypes by the mGluR5 antagonist MPEP. Neuropharmacology 2005;49:1053–66. [14] Boyle L, Kaufmann WE. The behavioral phenotype of FMR1 mutations. Am J Med Genet C Semin Med Genet 2010;154C:469–76. [15] Leteneur S, Gillet C, Sadeghi H, Allard P, Barbier F. Effect of trunk inclination on lower limb joint and lumbar moments in able men during the stance phase of gait. Clin Biomech 2009;24:190–5.
Contents lists available at ScienceDirect
Gait & Posture journal homepage: www.elsevier.com/locate/gaitpost
Short Communication
Implementation of a markerless motion analysis method to quantify hyperkinesis in males with fragile X syndrome Joan A. O’Keefe a, Alejandro A. Espinoza Orı´as b, Hassan Khan b, Deborah A. Hall c, Elizabeth Berry-Kravis c,d,e, Markus A. Wimmer a,b,* a
Department of Anatomy and Cell Biology, Rush University Medical Center, Chicago, IL 60612, United States Motion Analysis Laboratory, Department of Orthopedic Surgery, Rush University Medical Center, Chicago, IL 60612, United States Department of Neurological Sciences, Rush University Medical Center, Chicago, IL 60612, United States d Department of Pediatrics, Rush University Medical Center, Chicago, IL 60612, United States e Department of Biochemistry, Rush University Medical Center, Chicago, IL 60612, United States b c
A R T I C L E I N F O
A B S T R A C T
Article history: Received 21 November 2012 Received in revised form 14 October 2013 Accepted 19 October 2013
Hyperactive behavior – and implicitly, motion – in Fragile X syndrome (FXS) has been historically described using behavioral rating scales. While rating scales are the current standard outcome measures used in clinical research, they have limitations including their qualitative nature and subjectivity. The advent of new motion capture technologies has provided the opportunity to develop quantitative methods for measuring hyperactive motion. The hypotheses for this study were that a novel markerless motion analysis method (1) can quantitatively measure kinematic parameters, (2) can differentiate the level of hyperkinesis between control and FXS populations, and (3) will correlate with blind-reviewer synchronous video-capture methods and behavioral rating scale scores. Twenty young males (7-control, 13-FXS; ages 9–32) were studied using a standardized protocol in a markerless motion analysis suite. Behavioral scale questionnaires were filled out by parents and those scores were correlated with motion parameters (frequency and total traveled distance) of body segments during 30 s of story listening while standing in the observation space. The markerless system was able to track subjects and the two populations displayed significantly different quantities of motion, with larger amounts of motion in the FXS group (p < 0.05). Pearson’s correlation coefficients between frequency counts obtained from the markerless system and rater-based video capture were between 0.969 and 0.996 (p < 0.001). Significant correlations between rating scale scores and motion parameters ranged from 0.462 r 0.568 (p 0.040). These results suggest feasibility and validity of a markerless system as a non-invasive method able to quantify motion in individuals with hyperkinesis. ß 2013 Elsevier B.V. All rights reserved.
Keywords: Hyperkinesis Hyperactivity Fragile X syndrome Markerless motion analysis Behavioral rating scales
1. Introduction Fragile X syndrome (FXS) is the most common inherited cause of intellectual disability and the most common known genetic cause of autism [1]. It is associated with increased motor activity (hyperkinesis) that includes hyperactivity with increased body and limb movement and motor stereotypies (repetitive, non-goal directed movement patterns). Hyperactivity has been quantified in rodent models using open field tests to measure locomotor activity of the animal in a defined space [2]. An equivalent quantitative method does not exist for humans. Current outcome tools used
* Corresponding author at: Motion Analysis Laboratory, Rush University Medical Center, 1611 W Harrison Street, Suite 201 Orthopaedic Building, Chicago, IL 60612, United States. Tel.: +1 312 942 2789; fax: +1 312 942 2040. E-mail address: [email protected] (M.A. Wimmer). 0966-6362/$ – see front matter ß 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.gaitpost.2013.10.017
when investigating pharmaceutical (or other therapy) effects on hyperkinesis typically use parent/guardian rating scales [3,4]. These scales are the present gold standard for use in clinical trials research [5] and have validity and reliability [6,7]. However, they are not ideal because they are qualitative, inherently subjective, unable to detect small increments of behavioral changes, and prone to floor and/or ceiling effects. Traditional state of the art marker based motion analysis approaches to analyze hyperkinetic behaviors have not been reported in the FXS or autism literature, but they would likely have considerable limitations in this developmentally disabled population. Marker placement requires an expert spending a considerable amount of time with marker application and a subject who understands, does not object to removing clothing, and is cooperative with the instructions for motion analysis. A subject population with FXS and/or autism is expected to be non cooperative with the tedious marker application process due to behavioral issues including tactile
828
J.A. O’Keefe et al. / Gait & Posture 39 (2014) 827–830
defensiveness, anxiety, and lack of tolerability and understanding of the procedure. The above limitations in quantifying hyperkinesis and stereotypies, in combination with the critical need to quantify these core behavioral features of FXS and autism for efficacy research provided the impetus for the present study. Motion analysis is used in several fields of medical practice, including orthopedics, sports medicine, rehabilitation, and neurology. Accurate description of body kinematics is typically achieved with the use of reflective markers placed on anatomical landmarks to track limb, head and trunk position. Only recently have technological advances in image-recognition algorithms allowed development of systems robust enough to be used in clinical, sports and occupational biomechanics [8–11]. The objective of this study was to determine the feasibility of using markerless motion technology to quantify motor activity (specifically body mass center and limb segment movement) in young males with and without FXS. Our hypotheses are that a markerless motion analysis system will provide quantitative measures of hyperkinesis observed clinically in subjects with FXS, differentiate levels of motor activity between controls and those with FXS, and motion parameters of hyperkinesis will correlate with both videocaptured and clinical hyperactivity rating scale scores. Ultimately, we anticipate that this quantitative system will serve as a model for markerless motion measurement in hyperkinetic disorders. 2. Materials and methods The markerless motion capture system (Biostage, Organic Motion, New York, NY) acquired motion with fourteen video cameras mounted in a self-contained observation space bound by three reflective-canvas walls measuring with a footprint of 3.5 m 4.9 m and a wall height of 2.5 m (Fig. 1a). This system has been reported to have high correlation factors with a traditional marker-based system (VICONTM) for kinematic gait variables [12]. The 3D motion data was analyzed using The MotionMonitor1 inverse-dynamics analysis suite (Innsport, Chicago, IL). The following movement parameters are reported: (1) traveled distance of the body center-of-mass (COM), (2) motion frequency of the forearm and foot, and (3) traveled distance of the forearm and foot. Per convention, the right side was selected as the index body segment. Twenty male subjects including seven controls (age 9–22, mean 14.2 5.14 years) and thirteen FXS patients (age 10–32, mean 15.3 7.27 years) were recruited. The motion data was recorded during a standardized protocol. A variety of toys, manipulatives, and books spanning a broad age range of interest were offered to subjects
before actual motion recordings took place. Subjects were allowed to choose a favorite item before stepping into the Biostage area. Each subject had to go through a series of static standing postures (e.g. a Tpose during which the arms had to be held parallel to the ground for 2–5 s), to register the subject with the software and calibrate the system. Holding their favorite item, the subjects were then asked to listen to an audio-excerpt from Dr. Seuss’ ‘‘The Cat in the Hat’’ which has wide universal appeal to many age ranges. The audioclip was chosen because we empirically found it to hold the interest of all age ranges tested in this pilot study, including children with FXS who have intellectual difficulties and children and adult controls. The subjects were not instructed to stand still but were simply asked to listen to the story. The markerless system recorded motion data during three consecutive thirty-second intervals of story listening. Hyperactivity rating scales, including the Aberrant Behavior Checklist (ABC)-C – Hyperactivity Subscale score [3] and the Swanson, Nolan and Pelham IV (SNAP IV) – Hyperactivity/ Impulsivity subscale score [4] were filled out by parents on the day of testing and were correlated with the motion parameters in order to determine validity of the motion parameter outcomes. Further validity and reliability testing was done by having simultaneous, synchronized videotaped analysis of one control and one FXS subject’s motion during the markerless testing protocol repeated over three sessions. Concurrent criterion validity was tested by having 3 reviewers who were blind to the subject’s diagnosis perform off-line scoring of the frequency of eight motion parameters obtained from the video-capture method and then determining the correlations between these scores and those from the markerless system viewed on the TMM software interface. These included number of steps taken with the feet and arm movement in both upward and downward directions. Test–retest reliability of motion parameters was performed by examining the repeatability of the traveled distance measures obtained in the markerless system across the three sessions. The study was approved by the Rush University Institutional Review Board. All statistical analyses were performed using SPSS v 19.0 (SPSS Inc. Chicago, IL). Motion data from subjects were log-normalized before analysis. Independent student t-tests were then used for group comparisons. Correlations between the behavioral scales and the markerless kinematics parameters were obtained using two-tailed Spearman’s correlations. Two-tailed Pearson’s correlations coefficients were used to compare the kinetic data from the markerless system with the observer based counts from the video. Repeatability of the markerless motion parameters was determined using intraclass correlation coefficients. Statistical significance was established at a 0.05.
Fig. 1. (a) Depiction of the test area (Biostage). Three-dimensional view of the skeleton and center of body mass trajectory of representative, (b) FXS and (c) control subjects, respectively.
J.A. O’Keefe et al. / Gait & Posture 39 (2014) 827–830
829
Fig. 2. Traveled distances (in meters) of the (a) right forearm and (b) right foot relative to body mass center during 30 s of story listening. Note the differences in median and range between the control and FXS groups. The data are plotted on a logarithmic scale.
3. Results FXS subjects showed a large amount of movement in the Biostage. Fig. 1b and c show skeleton views of representative subjects from both experimental groups as viewed on the Motion Monitor’s interface. Arm and foot travel distances, were significantly greater in the FXS group compared to controls (p = 0.038 and 0.011, respectively, Fig. 2). The distance traveled by the body COM was greater in the FXS (3.3 4.0 m) versus control (1.1 0.5 m) groups. Due to the variable activity pattern within the FXS subjects, this difference did not reach significance (p = 0.177). All motion parameters were significantly associated with the ABC-C Hyperactivity subscale score (r = 0.481–0.568; p 0.032) and all except forearm frequency were significantly associated with the SNAP IVHyperactivity/Impulsivity Subscale score (r = 0.462–0.558; p 0.040; Table 1). Stronger correlations were obtained from travel distance parameters than motion frequency values. Motion parameters from the markerless system were significantly correlated with those counts obtained from the three blinded reviewers of simultaneous video-capture. The Pearson correlation coefficients ranged from 0.969 to 0.996 (p < 0.001). Repeated analysis of body segment movements generated similar total traveled distance values in all sessions. The 95% confidence interval of the intraclass correlation coefficients ranged from 0.919 to 0.986 (p < 0.001), thus providing further confidence in the repeatability of the method. 4. Discussion Kinematic assessment of body segments could be useful for future interventional studies attempting to quantify changes in hyperkinesis in patients with hyperkinetic disorders. This study is the first to attempt measurement of kinematic parameters obtained from a markerless motion capture system and correlate these measurements with scores on video-capture methodology
and behavioral rating scales in persons with FXS, a condition associated with hyperkinesis. The results obtained are encouraging because the method was able to discriminate FXS and control groups based on relative body segment travel. This was found to be a more sensitive hyperkinesis marker than ‘COM travel’, which did not reach significance, most likely due to low subject number and the variability of motion parameters within the FXS group. Clinically, FXS subjects move their arms frequently and/or repetitively. These are movements our control subjects never had. However, while standing, some control subjects swayed their upper body from one leg to the other, thus increasing the COM movement. This may have decreased the difference in COM travel distance between control and FXS groups to a non-significant value, despite the fact that many of the FXS subjects walked around excessively in the Biostage. The high correlations obtained between videotape counts from blinded raters and those from the markerless system demonstrate that we have excellent agreement between these two methodologies, suggesting validity of this new technology to quantify hyperkinesis. In addition, the traveled distance of body segments (a kinematic variable not computable from standard video analysis) showed high correlations with behavioral ratings of hyperkinesis, further strengthening the validity of this method in our subject population. While we did not provide proof of the accuracy of the measured distances, we demonstrated that these measures are reproducible and can be used to distinguish control from Fragile X subjects. Further, as mentioned already under Section 2, an earlier study found good agreement in kinematic measures between a marker-based and the markerless system, reporting correlation factors from 0.89 to 1.0 for the sagittal plane motion of the ankle, knee and hip joint [12]. The markerless motion system used in this study has important features and perhaps advantages over a traditional marker-based approach that made it feasible to measure hyperkinesis in persons with FXS. The system does not require the attachment of any leads
Table 1 Spearman correlation coefficients relating behavioral rating scores to motion parameters. Each table cell shows on the top row the correlation coefficient and below the significance value for a two-tailed correlation analysis. Scale/correlation coefficient
Forearm frequency
Foot frequency
Foot travel
Forearm travel
ABC-C (Hyperactivity)
0.533* 0.015
0.481* 0.032
0.568** 0.009
0.537* 0.015
SNAP-IV (Hyperactivity/Impulsivity)
0.330 0.155
0.462* 0.040
0.558* 0.011
0.508* 0.022
* **
Denotes significance at the 0.05 level (two-tailed). Denotes significance at the 0.01 level.
830
J.A. O’Keefe et al. / Gait & Posture 39 (2014) 827–830
or markers and the Biostage has minimal physical constraints, except for the requirement of the individual to stay in the testing area. These qualities increased the subject’s ability to participate in the testing protocol and minimized refusals to cooperate, thus allowing us to obtain quantitative results. Our experimental design was modeled after the open field hyperactivity measure utilized in behavioral studies of the Fragile X knockout mouse model of FXS [2,13,14]. In this test animals are placed in an enclosed area (the ‘‘open field’’) and their motion trajectory is tracked. The markerless motion capture system presented here provides similar behavioral analyses in humans, potentially facilitating translational research. It offers the additional advantage of quantifying the motion of specific body segments, and therefore is more relevant for tracking the manifestations of hyperkinesis in humans. While previous researchers have used video-capture methods synchronized to forceplate data in combination with an optical marker based approach to determine COM trajectory and velocity [15], this methodology requires significant cooperation and understanding on the part of the subject which would have been impossible in our developmentally disabled population. In summary, our feasibility study suggests that motion parameters obtained from the markerless system can quantify increased movement in subjects with FXS relative to controls. Preliminary findings demonstrate that hyperkinesis measures derived from this system are correlated with video-capture based methods and scores on standard behavioral checklists in both controls and individuals with FXS. These data require verification in larger samples with stringent reproducibility and clinical validity testing. Markerless motion capture of hyperkinesis may have important ramifications for removing current barriers in quantifying a major disruptive, phenotypic characteristic in FXS, autism, and other neurodevelopmental disorders. This technology has the potential to be used as a quantitative outcome measure for clinical research trials investigating methods to decrease hyperkinesis in these patient populations. Acknowledgements This study was supported by a pilot grant from the FRAXA Research Foundation (PI: Berry-Kravis). The authors also thank Mr. Robert Trombley from the Motion Analysis Lab at Rush University for his expert assistance with the motion analysis routines. Further,
we would like to thank Renee Kawecki, Chris Ferrigno, Gary Farkas, and Samir Chabra for their assistance with video analysis. Conflict of interest statement None of the authors have financial or other conflicts of interest in regards to this research. References [1] Wang LW, Berry-Kravis E, Hagerman RJ. Fragile X: leading the way for targeted treatments in autism. Neurotherapeutics 2010;7:264–74. [2] Min WW, Yuskaitis CJ, Yan Q, Sikorski C, Chen S, Jope RS, et al. Elevated glycogen synthase kinase-3 activity in Fragile X mice: key metabolic regulator with evidence for treatment potential. Neuropharmacology 2009;56:463–72. [3] Aman MG. Aberrant behavior checklist – community. East Aurora, New York: Slosson Educational Publications; 1994. [4] Swanson JM. School-based assessments and interventions for ADD students. Irvine, CA: KC Publishing; 1992. [5] Berry-Kravis E, Krause SE, Block SS, Guter S, Wuu J, Leurgans S, et al. Effect of CX516, an AMPA-modulating compound, on cognition and behavior in fragile X syndrome: a controlled trial. J Child Adolesc Psychopharmacol 2006;16:525–40. [6] Bihm EM, Poindexter AR. Cross-validation of the factor structure of the aberrant behavior checklist for persons with mental retardation. Am J Ment Retard 1991;96:209–11. [7] Miller ML, Fee VE, Jones CJ. Psychometric properties of ADHD rating scales among children with mental retardation. Res Dev Disabil 2004;25:477–92. [8] Corazza S, Mu¨ndermann L, Chaudhari AM, Demattio T, Cobelli C, Andriacchi TP. A markerless motion capture system to study musculoskeletal biomechanics: visual hull and simulated annealing approach. Ann Biomed Eng 2006;34:1019–29. [9] Moeslund TB, Hilton A, Kru¨ger V. A survey of advances in vision-based human motion capture and analysis. Comput Vision Image Understanding 2006;104:90–126. [10] Ray SJ, Teizer J. Real-time construction worker posture analysis for ergonomics training. Adv Eng Informatics 2012;26:439–55. [11] Urtasun R, Fleet DJ, Fua P. 3D people tracking with gaussian process dynamical models, IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Conf Proc 2006;1:238–45. [12] Oberla¨nder KD, Bru¨ggemann GP. Validation of a real-time markerless tracking system for clinical gait analysis ad-hoc results. In: 35th Annual Meeting of the American Society of Biomechanics. Conference Proceedings; 2011. Paper #304 (http://www.asbweb.org/conferences/2011/pdf/304.pdf; accessed on 13.10.13). [13] Yan QJ, Rammal M, Tranfaglia M, Bauchwitz RP. Suppression of two major Fragile X syndrome mouse model phenotypes by the mGluR5 antagonist MPEP. Neuropharmacology 2005;49:1053–66. [14] Boyle L, Kaufmann WE. The behavioral phenotype of FMR1 mutations. Am J Med Genet C Semin Med Genet 2010;154C:469–76. [15] Leteneur S, Gillet C, Sadeghi H, Allard P, Barbier F. Effect of trunk inclination on lower limb joint and lumbar moments in able men during the stance phase of gait. Clin Biomech 2009;24:190–5.