- Email: [email protected]
Perturbations on the base of support can be useful for gait recovery in children with CP?
S26
Abstracts / Gait & Posture 33S (2011) S1–S66 Table 1 RMSD values of two subjects for the angles of ankle plantar/dorsi-flexion (˛), knee flexion/extension (ˇ) and hip flexion/extension () intra-subject variability. Subject #1 (◦ )
RMSD RMSDV
Subject #2 (◦ )
˛
ˇ
˛
ˇ
5.2 24.0
3.4 26.5
7.2 5.0
6.4 4.1
4.2 2.3
7.6 0.3
Results Fig. 1a shows a subject image acquired by the non-infrared camera and Fig. 1b shows the same frame after segmentation and line-fitting. Table 1 shows the Root Mean Square Deviation (RMSD) of the two methods, and intra-subject variability (RMSDV ). Discussion Lower limb sagittal kinematics estimated from conventional and hybrid methods are comparable, making the latter a potential alternative for clinical use. Acknowledgement The authors thank BTS Bioengineering for the instrumentation and technical support provided for this study. References [1] Dobson F, Morris M, Baker R, Kerr Graham H. Gait Posture 2007;25:140–52. [2] Mündermann L, Corazza S, Andriacchi TP. J Neuroeng Rehabil 2006;3:6. [3] Stauffer C, Grimson WEL. IEEE Conf Comput Vision Pattern Recognit 1999;2:246–52. [4] Lam L, Seong-Whan L, Ching YS. IEEE Trans PAMI 1992;14(9):869–85.
doi:10.1016/j.gaitpost.2010.10.033 O29 Perturbations on the base of support can be useful for gait recovery in children with CP? M. Petrarca 1 , A. Colazza 1 , F. Patanè 1,2 , S. Rossi 1 , J.L. Jackson 1,2 , P. Cappa 1,2 , E. Castelli 1 1 Pediatric Neuro-Rehabilitation Division, Children’s Hospital “Bambino Gesù” IRCCS, Rome, Italy 2 Department of Mechanics and Aeronautics, “Sapienza” University of Rome, Italy
Fig. 1. Sample frame (a). Four axes fitted on the segmented frame (b).
acquisitions, Mixture of Gaussians (MoG) method [3] was used to subtract the moving body parts from the background. Afterwards, Medial Axis Transform (MAT) method was used to retrieve the thinned axis of the segmented images [4]. After the MAT, white colored underwear and socks were extracted from the original images via thresholding and the edge detection was applied to the extracted regions to find their intersection with the thinned axes. Then, the underwear and sock regions were overlapped on the resulting images which allowed to define three regions, two of which were considered rigid segments (pelvis and foot) and the other one composed by two rigid segments (thigh and shank). Next, to separate shank and thigh, anthropometric measures (segment length) were used in combination with the thinned axes resulting from the image processing and the knee location was identified.
Introduction The gait pattern of children with cerebral palsy is the consequence of central and peripheral contributions that cannot be really separated [1]. Recent studies [2,3] demonstrated that toe walk strategy shared the same resources both in children with cerebral palsy and in healthy subjects when they simulated toe walking. Biomechanical models seem to support this argument [4] hypothesising the energy conservation on the vertical plane. In order to clarify this hypothesis, we asked 10 children with CP to step on a compliant platform. Materials and methods 10 children with diplegia (9 ± 1 years old, 6 toe walkers and 4 no toe walkers) participated in this study. They walked barefoot on a 10 m walkway, unperturbed and perturbed by a compliant platform placed along the walkway. The platform was at the floor level and immediately before two AMTI force plates in order to allow consecutive steps on the platform and on the two force plates. The platform is robotized and controlled in order to move down under body weight. Gait analysis was performed by a Vicon Mx system with 8 cameras and a wireless surface EMG system. Three unperturbed and three perturbed gait analyses were conducted. The peaks of joints moments and powers were measured. For perturbed and unperturbed tests, stride was compared with similar
Abstracts / Gait & Posture 33S (2011) S1–S66
S27
Fig. 2. Mean data of the maximum powers and moments peak in four heel-walker children. Fig. 1. Mean of the powers and moments peak in six toe-walkers.
velocities in order to avoid gait speed effects. Statistical significance was tested with ANOVA. Results We analyzed kinetic unperturbed gait data and the stride after the platform perturbation. We analyzed powers and moments during stance. Although it was not statistically significant, when we compared perturbed and unperturbed gait, there was a tendency of increasing moments and powers during stance in the perturbed gait of toe walkers (Fig. 1). Ankle Power 1 and 2 refer to the power absorption and generation, respectively, characteristics of toe walkers in early stance, while Ankle Power 3 and 4 refer to the absorption and generation, respectively, common to all subjects in late stance. No toe walkers showed similar data in comparison of the two conditions (Fig. 2). EMG data of triceps, in Fig. 3a and b, showed, in comparison with free walk (dotted line), a reduction of early activities on the robotized platform (dashed line) and in the sequential stride on the force plates (dark line). Data also showed a high spread, sign of variability between subjects.
Discussion The differences between toe walkers and the other participants in the study showed changes of moments, powers and EMG activities during the stride executed on and after the step on the compliant platform. The scarce kinetics modifications between the two conditions respect to the EMG one, the latter similar in both groups as consequences of the perturbation, could be explained as the effect of the passive mechanical properties of soft tissues. Toe walkers seem to utilize a vertical conservation strategy exploiting the elastic properties of muscles and tendons. This strategy was characterized by the known repetition of absorption and generation of ankle power phases. During the perturbed stance, this mechanism was unable to work properly because of the compliant terrain. During the subsequent stride, it could be hypothesised that toe walkers tried to overcome this fault enhancing the bump. Compared to toe walkers, no toe walkers showed a more adaptive solution that seems not to require recovery after perturbation. Every attempt to modify toe walkers strategy should be anticipated by the study of the available resources on the single subject as indicated by the data spread. This is a preliminary observation and it will be necessary to study more subjects and to complete a comparison with healthy aged-matched children.
Fig. 3. (a) Diplegic toe walkers triceps enveloped EMG. (b) Diplegic no toe walkers triceps enveloped EMG.
S28
Abstracts / Gait & Posture 33S (2011) S1–S66
References [1] [2] [3] [4]
Crenna P. Neuroscience and Biobehavioral Reviews 1998;22(4):571–8. Perry J, et al. Archives of Physical Medicine and Rehabilitation 2003;84:7–13. Romkes J, et al. Gait & Posture 2007;26(4):577–86. Fonseca T, et al. Physical Therapy 2004;84(4):344–54.
doi:10.1016/j.gaitpost.2010.10.034 O30 A non invasive protocol to estimate muscle tendon lengths and moment arms through ultrasound images M.C. Bisi, F. Riva, R. Stagni DEIS, University of Bologna, Italy Introduction One of the important aspects of the application of muscleskeletal models in clinical diagnostic is the possibility to estimate muscle force contributions to joint moments, taking into account muscle co-contraction and physiological aspects of muscles. An accurate representation of the lower limb musculoskeletal system is required for the prediction of the muscle–tendon forces during human movement when using these models. The most recent muscle skeletal models are still sensitive to musculoskeletal geometry. Usually muscle–tendon lengths and moment arms are estimated from joint angles using published data obtained from cadaver specimens of different heights [1,2]. The determination of the length and of the line of action of the muscles personalized to subject morphology is one of the major steps in the development of reliable musculoskeletal models. In 2007, a non invasive method for determining the line of action of lower limb muscles with means of manual pointing on the subject has been proposed [3]. The aim of this study is to further improve the accuracy of these estimations using ultrasound images for identifying muscle insertions, origins and via points of a specific subject. Methods One healthy young subject [25 y,1.72 m, 61 kg] participated in the study. He was asked to perform an initial step exercise and then to continue walking at self selected speed while kinematic
data (SmartE, BTS, Milan, Italy) were collected. A 3 segment model of the subject right lower limb (thigh, leg and foot) was obtained. Gastrocnemius, soleus and tibialis anterior were selected as representative muscles of the lower limb. Origins (O), insertions (I) and via points of these muscles (details in Table 1) were calibrated using manual pointing [3] and using ultrasound images [4] while the participant was standing in neutral position. Retinaculum was calibrated both in neutral position and in complete dorsi-flexion of the ankle. These points were then reconstructed during the walk exercise. Muscle tendon lengths and moment arms were calculated through 5 different methods: (i) using equations taken from literature [1,2]; (ii) using IO points calibrated with manual pointing; (iii) using IO points calibrated with ultrasound images. In methods (ii) and (iii) tibialis anterior length and moment arm were calculated first with retinaculum calibrated with the ankle in neutral position (ii.a and iii.a) and then in complete dorsiflexion (ii.b and iii.b). Results Muscle tendon length variations obtained with the various methods were similar during the exercise. Offset values were underestimated or overestimated with the equation approach, while were comparable if obtained with kinematic or ultrasound IO calibrations. Moment arm results varied up to 40% of the mean values with the different estimation methods. In Fig. 1 exemplificative results for tibialis anterior during initial step and stance phases are shown. Tibialis anterior length and moment arm varied calibrating the retinaculum at different ankle position. Discussion Results of this study showed the importance of calibrating insertion, origin and via point of muscles in order to build reliable muscle skeletal models. Different methods showed similar trend of muscle tendon lengths during the exercise, while absolute value varied of about 5%. Moment arm results showed between method differences of 30–40% of medium muscle arm values, indicating how big errors can implicitly bring these inaccuracies to joint moment estimation in muscle-skeletal model predictions. Moreover variation of via point positions with movement (in this case of the retinaculum) can add errors in model estimations: further study
Table 1 Origins, insertions and via points of the selected muscles. Muscle
O
I
Via point
Tibialis anterior Gastrocnemius Soleus
Tuberosity tibiae middle point between Lateral and medial head from the lateral/medial condyl Posterior head of fibula
Tarsal–medial cuneiform Calcaneus–achilles tendon Calcaneus–achilles tendon
Retinaculum – –
Fig. 1. Tibialis anterior muscle–tendon length (a) and moment arms (b) obtained with different methods.
Abstracts / Gait & Posture 33S (2011) S1–S66 Table 1 RMSD values of two subjects for the angles of ankle plantar/dorsi-flexion (˛), knee flexion/extension (ˇ) and hip flexion/extension () intra-subject variability. Subject #1 (◦ )
RMSD RMSDV
Subject #2 (◦ )
˛
ˇ
˛
ˇ
5.2 24.0
3.4 26.5
7.2 5.0
6.4 4.1
4.2 2.3
7.6 0.3
Results Fig. 1a shows a subject image acquired by the non-infrared camera and Fig. 1b shows the same frame after segmentation and line-fitting. Table 1 shows the Root Mean Square Deviation (RMSD) of the two methods, and intra-subject variability (RMSDV ). Discussion Lower limb sagittal kinematics estimated from conventional and hybrid methods are comparable, making the latter a potential alternative for clinical use. Acknowledgement The authors thank BTS Bioengineering for the instrumentation and technical support provided for this study. References [1] Dobson F, Morris M, Baker R, Kerr Graham H. Gait Posture 2007;25:140–52. [2] Mündermann L, Corazza S, Andriacchi TP. J Neuroeng Rehabil 2006;3:6. [3] Stauffer C, Grimson WEL. IEEE Conf Comput Vision Pattern Recognit 1999;2:246–52. [4] Lam L, Seong-Whan L, Ching YS. IEEE Trans PAMI 1992;14(9):869–85.
doi:10.1016/j.gaitpost.2010.10.033 O29 Perturbations on the base of support can be useful for gait recovery in children with CP? M. Petrarca 1 , A. Colazza 1 , F. Patanè 1,2 , S. Rossi 1 , J.L. Jackson 1,2 , P. Cappa 1,2 , E. Castelli 1 1 Pediatric Neuro-Rehabilitation Division, Children’s Hospital “Bambino Gesù” IRCCS, Rome, Italy 2 Department of Mechanics and Aeronautics, “Sapienza” University of Rome, Italy
Fig. 1. Sample frame (a). Four axes fitted on the segmented frame (b).
acquisitions, Mixture of Gaussians (MoG) method [3] was used to subtract the moving body parts from the background. Afterwards, Medial Axis Transform (MAT) method was used to retrieve the thinned axis of the segmented images [4]. After the MAT, white colored underwear and socks were extracted from the original images via thresholding and the edge detection was applied to the extracted regions to find their intersection with the thinned axes. Then, the underwear and sock regions were overlapped on the resulting images which allowed to define three regions, two of which were considered rigid segments (pelvis and foot) and the other one composed by two rigid segments (thigh and shank). Next, to separate shank and thigh, anthropometric measures (segment length) were used in combination with the thinned axes resulting from the image processing and the knee location was identified.
Introduction The gait pattern of children with cerebral palsy is the consequence of central and peripheral contributions that cannot be really separated [1]. Recent studies [2,3] demonstrated that toe walk strategy shared the same resources both in children with cerebral palsy and in healthy subjects when they simulated toe walking. Biomechanical models seem to support this argument [4] hypothesising the energy conservation on the vertical plane. In order to clarify this hypothesis, we asked 10 children with CP to step on a compliant platform. Materials and methods 10 children with diplegia (9 ± 1 years old, 6 toe walkers and 4 no toe walkers) participated in this study. They walked barefoot on a 10 m walkway, unperturbed and perturbed by a compliant platform placed along the walkway. The platform was at the floor level and immediately before two AMTI force plates in order to allow consecutive steps on the platform and on the two force plates. The platform is robotized and controlled in order to move down under body weight. Gait analysis was performed by a Vicon Mx system with 8 cameras and a wireless surface EMG system. Three unperturbed and three perturbed gait analyses were conducted. The peaks of joints moments and powers were measured. For perturbed and unperturbed tests, stride was compared with similar
Abstracts / Gait & Posture 33S (2011) S1–S66
S27
Fig. 2. Mean data of the maximum powers and moments peak in four heel-walker children. Fig. 1. Mean of the powers and moments peak in six toe-walkers.
velocities in order to avoid gait speed effects. Statistical significance was tested with ANOVA. Results We analyzed kinetic unperturbed gait data and the stride after the platform perturbation. We analyzed powers and moments during stance. Although it was not statistically significant, when we compared perturbed and unperturbed gait, there was a tendency of increasing moments and powers during stance in the perturbed gait of toe walkers (Fig. 1). Ankle Power 1 and 2 refer to the power absorption and generation, respectively, characteristics of toe walkers in early stance, while Ankle Power 3 and 4 refer to the absorption and generation, respectively, common to all subjects in late stance. No toe walkers showed similar data in comparison of the two conditions (Fig. 2). EMG data of triceps, in Fig. 3a and b, showed, in comparison with free walk (dotted line), a reduction of early activities on the robotized platform (dashed line) and in the sequential stride on the force plates (dark line). Data also showed a high spread, sign of variability between subjects.
Discussion The differences between toe walkers and the other participants in the study showed changes of moments, powers and EMG activities during the stride executed on and after the step on the compliant platform. The scarce kinetics modifications between the two conditions respect to the EMG one, the latter similar in both groups as consequences of the perturbation, could be explained as the effect of the passive mechanical properties of soft tissues. Toe walkers seem to utilize a vertical conservation strategy exploiting the elastic properties of muscles and tendons. This strategy was characterized by the known repetition of absorption and generation of ankle power phases. During the perturbed stance, this mechanism was unable to work properly because of the compliant terrain. During the subsequent stride, it could be hypothesised that toe walkers tried to overcome this fault enhancing the bump. Compared to toe walkers, no toe walkers showed a more adaptive solution that seems not to require recovery after perturbation. Every attempt to modify toe walkers strategy should be anticipated by the study of the available resources on the single subject as indicated by the data spread. This is a preliminary observation and it will be necessary to study more subjects and to complete a comparison with healthy aged-matched children.
Fig. 3. (a) Diplegic toe walkers triceps enveloped EMG. (b) Diplegic no toe walkers triceps enveloped EMG.
S28
Abstracts / Gait & Posture 33S (2011) S1–S66
References [1] [2] [3] [4]
Crenna P. Neuroscience and Biobehavioral Reviews 1998;22(4):571–8. Perry J, et al. Archives of Physical Medicine and Rehabilitation 2003;84:7–13. Romkes J, et al. Gait & Posture 2007;26(4):577–86. Fonseca T, et al. Physical Therapy 2004;84(4):344–54.
doi:10.1016/j.gaitpost.2010.10.034 O30 A non invasive protocol to estimate muscle tendon lengths and moment arms through ultrasound images M.C. Bisi, F. Riva, R. Stagni DEIS, University of Bologna, Italy Introduction One of the important aspects of the application of muscleskeletal models in clinical diagnostic is the possibility to estimate muscle force contributions to joint moments, taking into account muscle co-contraction and physiological aspects of muscles. An accurate representation of the lower limb musculoskeletal system is required for the prediction of the muscle–tendon forces during human movement when using these models. The most recent muscle skeletal models are still sensitive to musculoskeletal geometry. Usually muscle–tendon lengths and moment arms are estimated from joint angles using published data obtained from cadaver specimens of different heights [1,2]. The determination of the length and of the line of action of the muscles personalized to subject morphology is one of the major steps in the development of reliable musculoskeletal models. In 2007, a non invasive method for determining the line of action of lower limb muscles with means of manual pointing on the subject has been proposed [3]. The aim of this study is to further improve the accuracy of these estimations using ultrasound images for identifying muscle insertions, origins and via points of a specific subject. Methods One healthy young subject [25 y,1.72 m, 61 kg] participated in the study. He was asked to perform an initial step exercise and then to continue walking at self selected speed while kinematic
data (SmartE, BTS, Milan, Italy) were collected. A 3 segment model of the subject right lower limb (thigh, leg and foot) was obtained. Gastrocnemius, soleus and tibialis anterior were selected as representative muscles of the lower limb. Origins (O), insertions (I) and via points of these muscles (details in Table 1) were calibrated using manual pointing [3] and using ultrasound images [4] while the participant was standing in neutral position. Retinaculum was calibrated both in neutral position and in complete dorsi-flexion of the ankle. These points were then reconstructed during the walk exercise. Muscle tendon lengths and moment arms were calculated through 5 different methods: (i) using equations taken from literature [1,2]; (ii) using IO points calibrated with manual pointing; (iii) using IO points calibrated with ultrasound images. In methods (ii) and (iii) tibialis anterior length and moment arm were calculated first with retinaculum calibrated with the ankle in neutral position (ii.a and iii.a) and then in complete dorsiflexion (ii.b and iii.b). Results Muscle tendon length variations obtained with the various methods were similar during the exercise. Offset values were underestimated or overestimated with the equation approach, while were comparable if obtained with kinematic or ultrasound IO calibrations. Moment arm results varied up to 40% of the mean values with the different estimation methods. In Fig. 1 exemplificative results for tibialis anterior during initial step and stance phases are shown. Tibialis anterior length and moment arm varied calibrating the retinaculum at different ankle position. Discussion Results of this study showed the importance of calibrating insertion, origin and via point of muscles in order to build reliable muscle skeletal models. Different methods showed similar trend of muscle tendon lengths during the exercise, while absolute value varied of about 5%. Moment arm results showed between method differences of 30–40% of medium muscle arm values, indicating how big errors can implicitly bring these inaccuracies to joint moment estimation in muscle-skeletal model predictions. Moreover variation of via point positions with movement (in this case of the retinaculum) can add errors in model estimations: further study
Table 1 Origins, insertions and via points of the selected muscles. Muscle
O
I
Via point
Tibialis anterior Gastrocnemius Soleus
Tuberosity tibiae middle point between Lateral and medial head from the lateral/medial condyl Posterior head of fibula
Tarsal–medial cuneiform Calcaneus–achilles tendon Calcaneus–achilles tendon
Retinaculum – –
Fig. 1. Tibialis anterior muscle–tendon length (a) and moment arms (b) obtained with different methods.