A single gyrometer inside an instrumented running shoe allows mobile determination of gait cycle and pronation velocity during outdoor running Torsten Brauner, Doris Oriwol, Thorsten Sterzing, Thomas L. Milani Department of Human Locomotion, Chemnitz University of Technology, Chemnitz, Germany Introduction Excessive pronation is discussed to be a factor in the development of overuse injuries. Prolonged running has been claimed to increase pronation and pronation velocity (van Gheluwe & Madsen 1997, Derrick et al. 2002). However, this coherence is contradicted by findings of (Sterzing & Hennig 1999, Butler et al. 2007). A reason for this discrepancy might lie within the different methodical study designs used, because investigations of fatiguing effects on pronation were either bond to treadmill running, to pre-post-measurements, or were biased due to, although mobile, but unpractical goniometer measurements. A practical and reliable device for mobile pronation measurement is missing so far. A lot of research has been put into the development of devices for mobile kinematic measurements. Knee joint kinematics can be determined with high accuracy using gyrometers and accelerometers (Favre et al., 2008). For ankle kinematics during running, the use of accelerometer is highly problematic due to the massive crosstalk of the impact (Takeda et al 2009). However, in a laboratory study we were able to demonstrate that the sole use of a gyrometer at the foot resulted in adequate pronation measurements in the frontal plane (Brauner et al. 2009). The purpose of this study was to investigate whether the use of a single gyrometer integrated directly inside the midsole of a running shoe is sufficient to detect the gait cycle and to determine maximum pronation velocity during regular outdoor running. Methods Instrumented shoe A gyrometer (Murata ENC-03R, Murata Manufacturing Company, Ltd., Japan) was fixed with an elastic adhesive inside the midsole of a regular running shoe. The axis of the sensor was aligned with the longitudinal axis of the shoe and the sensor was positioned approximately 15mm from the posterior end of the shoe, 15mm laterally, and 20mm above the floor. A cable connected the sensor to a lightweight (380g) data logger (eLAS.net MultiLOG2, MSR-Electronics, Switzerland) secured in a waist belt where the signal was recorded at 1 kHz. Gyrometer data was high-pass filtered at 0.1Hz (1st order Butterworth) to eliminate sensor drift prior to further processing. Algorithm development Data of one subject (S1: 27yrs, 83.3kg, 182cm) was used to develop an algorithm for mathematically detection of the gait cycle and determination of maximum pronation velocity. For that purpose, S1 performed running trials in the laboratory at three different running speeds where gyrometer data of the instrumented shoe and ground reaction forces (Kistler 9287BA, 1 kHz) were recorded simultaneously. Thereby, it was possible to compare gait characteristics determined by GRF data with the gyrometer signal (Figure 1-a). The gyrometer signal was low-pass filtered at 6Hz (1st order Butterworth) to reduce the signal to the main gait cycle characteristics only. Prior to ground contact the filtered gyrometer signal showed a very prominent high inversion velocity peak (VELmin) that could be identified clearly. Maximum pronation velocity (VELmax) was determined as the maximum value of the original gyro signal between VELmin and the next minimum (in Figure 1-b). Algorithm validation A within-subject validation was performed with data collected during a 16-day relay run around Germany (Heidenfelder et al. 2009). During this event S1 performed 27 consecutive runs (57.2±20.4min, 12.1±3.3km) with an 11h break between runs. He wore the instrumented shoe at each run during that event. VELmax were identified by applying the developed algorithm on gyrometer data. Afterwards, results were visually verified and detection errors counted. False detections as well as ignoring an obvious VELmax peak were counted as errors.
For between-subject validation ten injury-free male subjects (23.5±4.0yrs, 86.5±13.0kg, 185±6 cm) performed outdoor running trials (27±7mins, 1981±585steps) with the same instrumented shoe. The algorithm was applied on their gyrometer data and the results were visually verified as well.
Results & Discussion Gyrometer signals demonstrated a very stable scheme between runs of one subject as well as between subjects. Hence, satisfying results in gait detection and VELmax determination were achieved. Within the 105,623 analyzed steps the overall error rate was with 0.19% acceptable low. Neither the within validation (91 errors in 85810 steps) nor the between-subject validation (114 errors in 15,170 steps) revealed considerable deficiencies of the algorithm. One subject showed a high error rate of 4.6% due to false detection of VELmax peaks. A threshold of 150°/s for VELmax detection would reduce this problem considerably. Conclusion A single gyrometer fixed inside the shoe proved to be reliable in determining gait cycle and maximum pronation velocity in outdoor running. Furthermore, the algorithm is able to detect the first steps at the beginning of a run and works thereby automatically. Hence, a system has been developed to study effects of fatigue on pronation in a natural running environment. Further progress in the ability to determine absolute angles is needed as it would enhance mobile pronation measurements considerably. References Brauner, T. et al. (2009). Submitted to ISB conference, Cape Town, South Africa. Butler, R. et al. (2007). Gait & Posture, 26(2): 219–225. Derrick, T. R. et al. (2002). Medicine and science in sports and exercise, 34(6): 998–1002. Favre, J. et al. (2008). Journal of Biomechanics, 41(5): 1029–1035. Gheluwe van, B., & Madsen Claire (1997). Journal of Applied Biomechanics, 13(1): 66–75. Heidenfelder, J., et al. (2009). Submitted to ISB conference, Cape Town, South Africa. Sterzing, T., & Hennig, E. M. (1999). Proceedings of the 4th Footwear Symp., Canmore, Canada: 88-89. Takeda, R., et al. (2009). Journal of Biomechanics, 42(3): 223–233. Acknowledgements This research was supported by PUMA Inc., Germany.
A single gyrometer inside an instrum A single gyrometer inside an instrum mented running shoe allows mobile mented running shoe allows mobile d t dete determination of gait cycle and prona i attio of gaitt cycle a d p o ation velocity during outdoor running at atio elocitty du d i g outdoo td u i g Torsten Brauner Brauner, Doris Oriwol Oriwol, Thorrsten Sterzing Sterzing, Thomas L L. Milani Algorithm validation (between‐subjects) Al ith lid ti (b t bj t )
I Introduction d i Excessive pronation is discussed to be a factor in the development of overuse injuries Prolonged running has been claimed to increase pronation and pronation injuries. velocity l ( (van Gheluwe h l & Madsen 1997, Derrickk et al.l 2002). ) However, this h coherence is contradicted by findings of (Sterzing & Hennig 1999, 1999 Butler et al. al 2007) A reason for f this thi discrepancy di i ht lie li within ithi the th different diff t methodical th di l 2007). might studyy designs g used,, because investigations g of fatiguing g g effects on p pronation were running to pre‐post‐measurements, pre post measurements or were biased due either bond to treadmill running, to, although measurements. A practical lh h mobile, bil but b unpractical i l goniometer i i l and d reliable device for mobile pronation measurement is missing so far. far A lot of research has been put into the development of devices for mobile kinematic measurements. Knee joint j Kinematics can be determined with high accuracy using gyrometers t and d accelerometers l t (Favre (F ett al., l 2008). 2008) For F ankle kl Kinematics during running, the use of accelerometer is highly Problematic due to the massive crosstalk of the impact ((Takeda et al 2009). ) However,, in a laboratoryy studyy we were able to demonstrate that the sole use of a gyrometer at the f t resulted lt d in i adequate d t pronation ti measurements t in i the th foot frontal plane (Brauner et al. 2009). The h purpose off this h study was to investigate whether h h the h use of a single gyrometer integrated directly inside the midsole of a running i shoe h is i sufficient ffi i t to t detect d t t the th gait it cycle l and d to t determine maximum p pronation velocityy duringg regular g outdoor running. running Fig. 1: S1 on the road Fi 1 S1 th d
• Ten injury‐free male subjects (S2‐S11, (S2‐S11 23.5 23 5 ± 4.0yrs, 4 0yrs 86.5 86 5 ± 13.0kg, 13 0kg 185 ± 6cm) performed outdoor runningg trials ((27 ± 7mins,, 1981 ± 585steps) p p) • Same instrumented shoe • Algorithm was applied on their gyrometer data, the results were visually verified ifi d and d errors counted d
R lt & Di i Results & Discussion Algorithm validation (within subject) Algorithm validation (within‐subject) • Gyrometer signals demonstrated a very stable scheme G t i l d t t d t bl h • 91 errors in 85810 steps Æ 91 errors in 85810 steps Æ Error rate of 0.11% Error rate of 0 11% • Satisfying results in gait detection and VELmax y g g determination for S1
Fi 4 E Fig. 4: Exemplary data of 200 steps of S1 (blue line) and VELmax l d t f 200 t f S1 (bl li ) d VEL ( d li ) (red line)
200 steps 200 steps
Al ith Algorithm validation (between‐subjects) lid ti (b t bj t )
M h d Methods Instrumented shoe Instrumented shoe • Gyrometer (ENC‐03R, Murata Manufacturing G t (ENC 03R M t M f t i Company Ltd Japan) Company, Ltd., Japan) • Fixed with an elastic adhesive inside the midsole of a regular running shoe regular running shoe • Axis of the sensor was aligned with the longitudinal Axis of the sensor was aligned with the longitudinal Fig. 2: Instrumented shoe Fig 2: Instrumented shoe axis of the shoe f h h • Data logger (eLAS.net MultiLOG2, MSR‐Electronics, Switzerland, 380g) Data logger (eLAS net MultiLOG2 MSR Electronics Switzerland 380g) • Data recorded at 1 kHz Data recorded at 1 kHz • High‐pass filter (0.1Hz , 1 Hi h filt (0 1H 1st order, Butterworth) to eliminate sensor drift d B tt th) t li i t d ift
• 114 114 errors in 15,170 steps Æ errors in 15 170 steps Æ Error rate of 0.75% Error rate of 0 75% • One subject showed a high error rate of 4.6% due to false detection of VELmax j g peaks peaks • A threshold of 150 A threshold of 150°/s /s for VELmax for VELmax detection would reduce this problem detection would reduce this problem considerably d bl • Satisfying results in gait detection and VELmax Satisfying results in gait detection and VELmax determination over all subjects determination over all subjects S2 S3 S4
Algorithm development (gait cycle & maximum pronation velocity d determination) i i )
S5
• One subject (S1: 27yrs, 27yrs 83.3kg, 83 3kg 182cm) • Laboratoryy test with Gyrometer y and Force Plate at different runningg velocities • For F Gait G it Cycle C l Detection D t ti gyrometer t signal i l was low‐pass l filt d att 6Hz filtered 6H (1st order Butterworth) • Prior to gground contact the filtered gy gyrometer signal g showed a veryy p prominent high inversion velocity peak (Figure 3 – yellow dots) • Maximum pronation velocity (VELmax) was determined as the maximum value off the th original i i l gyro signal i l between b t VEL i and VELmin d the th nextt minimum i i
S6 S7 S8 Fig. 5: Exemplary gyrometer data of approximately 15 steps from S2‐S8 g p y gy pp y p
Conclusion A single i l gyrometer fixed fi d inside i id the h shoe h proved d to be b reliable li bl in i determining d i i gait i cycle and maximum pronation velocity in outdoor running. Furthermore, the algorithm is able to detect the first steps at the beginning of a run and works therebyy automatically. y Hence,, a system y has been developed p to studyy effects of fatigue on pronation in a natural running environment. environment F th progress in Further i the th ability bilit to t determine d t i absolute b l t angles l is i needed d d as it would ld enhance mobile pronation measurements considerably.
Fig. 3: Original gyrometer data (black line), 6Hz. filtered signal (blue line), Fi 3 O i i l t d t (bl k li ) 6H filt d i l (bl li ) VELmax (red line) and Minima (yellow and blue dots) ( ) (y )
5 5 steps steps
Al ith Algorithm validation (within‐subject) lid ti ( ithi bj t) • Data Data was collected during a 16‐day relay run around Germany was collected during a 16‐day relay run around Germany (Heidenfelder et al. 2009)) • S1 performed 27 consecutive runs (57.2 ± S1 performed 27 consecutive runs (57 2 ± 20.4min, 12.1 ± 20 4min 12 1 ± 3.3km) with 3 3km) with instrumented shoe, 11h break between runs instrumented shoe 11h break between runs • VELmax were identified by applying the developed algorithm on gyrometer d f db l h d l d l h data • Results were visually verified and detection errors counted Results were visually verified and detection errors counted • False detections as well as ignoring an obvious VELmax F l d t ti ll i i b i VEL peak were counted as k t d errors errors
Torsten Brauner torsten brauner@phil tu chemnitz de
[email protected] +49 371 531 36615
References Brauner, T. et al. (2009). ISB conference, Cape Town, South Africa. Brauner, T. et al. (2009). ISB conference, Cape Town, South Africa. Butler R et al (2007) Gait & Posture 26(2): 219 225 Butler, R. et al. (2007). Gait & Posture, 26(2): 219–225. Derrick, T. R. et al. (2002). Med & Sci , ( ) in Sports & Exec, 34(6): 998–1002. p , ( ) Favre J et al (2008) J Biomech 41(5): 1029–1035 Favre, J. et al. (2008). J Biomech, 41(5): 1029 1035. Gh l Gheluwe van, B., & Madsen Claire (1997). J. Appl B &M d Cl i (1997) J A l Biomech, 13(1): 66–75. Bi h 13(1) 66 75 Heidenfelder, J., et al. (2009. ISB conference, Cape Town, South Africa. Heidenfelder, J., et al. (2009. ISB conference, Cape Town, South Africa. Sterzing, T., & Hennig, E. M. (1999). Proceedings of the 4th Footwear Symp., Sterzing T & Hennig E M (1999) Proceedings of the 4th Footwear Symp Canmore, Canada: 88‐89. Takeda R et al (2009) J Biomech 42(3): 223 233 Takeda, R., et al. (2009). J. Biomech, 42(3): 223–233.
Chemnitz Univerrsity of Technology Department of Human Locomotion 09107 Chemnitz, it Germany G
Acknowledgements Thi This research was supported by PUMA Inc., Germany. h t d b PUMA I G XXII Congress of the International Society of Biomechanics July 5th - 9th, 9th 2009 | Cape Town (South Africa)
