Extended Abstract (for invited Faculty only) Biomechanics

11.2.2 - Clinical Gait Analysis: A Toolbox to Study Dynamic Joint Loading

Presentation Topic
Biomechanics
Date
13.04.2022
Lecture Time
15:00 - 15:15
Room
Potsdam 1
Session Type
Special Session
Speaker
  • B. Horsak (St. Pölten, AT)
Authors
  • B. Horsak (St. Pölten, AT)

Abstract

Introduction

The quantification and analysis of human movement have always challenged researchers and clinicians over the past decades. While it was a time-consuming and cumbersome task in the early days of motion capturing, technical advancements such as increasing processing power, (wearable) sensor technologies, the ability to collect and process large amount of data, and machine learning approaches are revolutionizing clinical gait analysis (CGA) for both clinical and research settings. This talk aims to offer a brief overview of state-of-the-art methods to quantify human locomotion three-dimensionally with special emphasis on estimating internal (knee) joint loading. The methods covered will encompass standard marker-based CGA which is typically used in clinical settings, to more complex physics-based musculoskeletal simulations. Finally, this talk provides a brief perspective on how the rapid technological development will shape the future of estimating dynamic (knee) joint loading not only in laboratory settings but also in the field.

Content

To quantify human movement three-dimensionally the current technology used in CGA are infrared-based motion capture systems. A motion capture system comprises several synchronized infrared cameras which record the three-dimensional coordinates of small reflective spherical markers that are attached with adhesive tape to the patient's skin above anatomical landmarks. These coordinates, together with a kinematic biomechanical model, then allow the calculation of joint kinematics such as knee joint angles [1]. During laboratory gait analysis, force plates are used to measure the occurring ground reaction forces. When these forces are combined with the kinematic information, an inverse-dynamic approach can be used to estimate joint reaction forces and joint moments (kinetic variables), e.g., the internal knee flexion moment. The medical history and a rigorous clinical examination are then used together with the CGA data to inform the medical diagnosis, assessment, monitoring, or outcome prediction for both orthopedic and neuro-orthopedic patients [2].

Although in-vivo joint loading cannot be measured directly with motion capturing, a commonly used surrogate to estimate internal medial knee joint loading is the frontal knee adduction moment (KAM). The rationale for this surrogate is the association between higher KAM and observed radiographic degeneration of the knee structure and cartilage [3,4]. However, while research has demonstrated that biofeedback-assisted gait retraining can reduce KAM and pain as well as improve function in knee osteoarthritis patients [5,6], it also provides evidence that KAM is only moderately associated with the actual medial knee joint contact force [7]. Thus, while KAM is simple to assess, it only provides a moderately reasonable surrogate for internal knee joint loading.

Physics-based simulations expand upon common CGA methodology by utilizing more complex biomechanical models [8]. Basically, they provide a mathematical description of the human (neuro-) musculoskeletal system. Their main advantage is that they can be used to estimate internal structure loading, such as of the ligaments, or muscle forces, as well as joint contact loading. Thereby allowing a more in-depth understanding of the internal loading during human movement. Further, these models offer the potential of personalization by adjusting, for instance, the muscle geometry or bone-cartilage (knee) morphology of the model based on medical images of a patient [9]. While these physics-based (and personalized) simulations typically need substantial processing power, recent research has demonstrated that they can be run in real-time to offer biofeedback about the medial tibiofemoral contact force or Achilles tendon strain during gait retraining [10,11]. Physics-based simulations are becoming increasingly popular in both research and clinical settings.

Although musculoskeletal simulations can support clinical decision making [8] the need for highly trained personnel and associated hardware and time requirements lead to a reduced accessibility especially in clinical settings [12]. However, the ongoing rapid technological development might offer great potential to make these methods more accessible in the future. One prominent example is Machine and Deep Learning (ML). ML allows the bypassing of several bottleneck calculations in physics-based musculoskeletal modeling workflows. To this end, artificial neuronal networks (ANN) are trained with a large amount of motion capture data and musculoskeletal simulation results. After successful training, they can estimate internal joint or muscle forces in a fraction of the time necessary to run traditional musculoskeletal simulations. Current approaches seem promising and demonstrate high correlations between ML-based predicted internal forces and actual simulation results [13–15]. Recently, it was also shown that these approaches can be based on data from wearable inertial measurements units (IMUs) instead of motion capture data [16,17] thereby offering the potential to bring musculoskeletal simulations into the field. Another very recent approach which utilizes ML is marker-less three-dimensional human pose estimation from either monocular or multi-view settings [18]. This technology allows the estimation of human movement two- and three-dimensionally based simply on video images without the need for markers being attached to the skin. The feasibility of such an approach to measure for example the peak KAM based only on positions obtainable from two-dimensional video analysis was recently demonstrated [19]. Marker-less motion capture systems, to date, are already partly commercially available and allow to estimate three-dimensional gait kinematics with errors of less than 5.5 degrees with the exception of segmental rotations about the longitudinal axis [20]. While error rates of such approaches presently only seem partly satisfying, marker-less motion capturing techniques are currently under heavy development and presumably will become increasingly more powerful over time.

In conclusion, CGA offers various approaches to investigate dynamic (knee) joint loading, and recent research demonstrates that these predominantly laboratory-based measurements and simulations might soon be even available outside of the laboratory with the support of wearable sensors and artificial intelligence. The next step will be to combine these technologies into well-designed and easy to use products that can support clinicians and patients during medical daily routine [21].

References

[1] R. Baker, Measuring walking: a handbook of clinical gait analysis, Mac Keith Press, London, 2013.

[2] R. Baker, Gait analysis methods in rehabilitation, J. NeuroEngineering Rehabil. 3 (2006) 4. https://doi.org/10.1186/1743-0003-3-4.

[3] K.L. Bennell, K.-A. Bowles, Y. Wang, F. Cicuttini, M. Davies-Tuck, R.S. Hinman, Higher dynamic medial knee load predicts greater cartilage loss over 12 months in medial knee osteoarthritis, Ann. Rheum. Dis. 70 (2011) 1770–1774. https://doi.org/10.1136/ard.2010.147082.

[4] E.F. Chehab, J. Favre, J.C. Erhart-Hledik, T.P. Andriacchi, Baseline knee adduction and flexion moments during walking are both associated with 5 year cartilage changes in patients with medial knee osteoarthritis, Osteoarthritis Cartilage. 22 (2014) 1833–1839. https://doi.org/10.1016/j.joca.2014.08.009.

[5] J. Bowd, P. Biggs, C. Holt, G. Whatling, Does Gait Retraining Have the Potential to Reduce Medial Compartmental Loading in Individuals With Knee Osteoarthritis While Not Adversely Affecting the Other Lower Limb Joints? A Systematic Review, Arch. Rehabil. Res. Clin. Transl. 1 (2019) 100022. https://doi.org/10/gnt2s6.

[6] L.C. Pereira, J. Runhaar, J. Favre, B.M. Jolles, S. Bierma-Zeinstra, Association between changes in the knee adduction moment and changes in knee pain and function in response to non-surgical biomechanical interventions for medial knee osteoarthritis: a systematic review, Eur. J. Phys. Rehabil. Med. (2021). https://doi.org/10/gnt2s7.

[7] R.E. Richards, M.S. Andersen, J. Harlaar, J.C. van den Noort, Relationship between knee joint contact forces and external knee joint moments in patients with medial knee osteoarthritis: effects of gait modifications, Osteoarthritis Cartilage. 26 (2018) 1203–1214. https://doi.org/10/gd4jhh.

[8] B.A. Killen, A. Falisse, F. De Groote, I. Jonkers, In Silico-Enhanced Treatment and Rehabilitation Planning for Patients with Musculoskeletal Disorders: Can Musculoskeletal Modelling and Dynamic Simulations Really Impact Current Clinical Practice?, Appl. Sci. 10 (2020) 7255. https://doi.org/10/gnkdhk.

[9] C.R. Smith, R.L. Lenhart, J. Kaiser, M.F. Vignos, D.G. Thelen, Influence of Ligament Properties on Tibiofemoral Mechanics in Walking, J. Knee Surg. 29 (2016) 99–106. https://doi.org/10.1055/s-0035-1558858.

[10] C. Pizzolato, M. Reggiani, D.J. Saxby, E. Ceseracciu, L. Modenese, D.G. Lloyd, Biofeedback for Gait Retraining Based on Real-Time Estimation of Tibiofemoral Joint Contact Forces, IEEE Trans. Neural Syst. Rehabil. Eng. 25 (2017) 1612–1621. https://doi.org/10.1109/TNSRE.2017.2683488.

[11] C. Pizzolato, V.B. Shim, D.G. Lloyd, D. Devaprakash, S.J. Obst, R. Newsham-West, D.F. Graham, T.F. Besier, M.H. Zheng, R.S. Barrett, Targeted Achilles Tendon Training and Rehabilitation Using Personalized and Real-Time Multiscale Models of the Neuromusculoskeletal System, Front. Bioeng. Biotechnol. 8 (2020). https://doi.org/10.3389/fbioe.2020.00878.

[12] W.S. Burton, C.A. Myers, P.J. Rullkoetter, Machine learning for rapid estimation of lower extremity muscle and joint loading during activities of daily living, J. Biomech. 123 (2021) 110439. https://doi.org/10/gkgk44.

[13] L. Rane, Z. Ding, A.H. McGregor, A.M.J. Bull, Deep Learning for Musculoskeletal Force Prediction, Ann. Biomed. Eng. 47 (2019) 778–789. https://doi.org/10/gf3f68.

[14] M.M. Ardestani, Z. Chen, L. Wang, Q. Lian, Y. Liu, J. He, D. Li, Z. Jin, Feed forward artificial neural network to predict contact force at medial knee joint: Application to gait modification, Neurocomputing. 139 (2014) 114–129. https://doi.org/10/gh3884.

[15] T.T. Dao, From deep learning to transfer learning for the prediction of skeletal muscle forces, Med. Biol. Eng. Comput. 57 (2019) 1049–1058. https://doi.org/10/ghhvsp.

[16] M. Mundt, W.R. Johnson, W. Potthast, B. Markert, A. Mian, J. Alderson, A Comparison of Three Neural Network Approaches for Estimating Joint Angles and Moments from Inertial Measurement Units, Sensors. 21 (2021) 4535. https://doi.org/10/gmn54p.

[17] A. De Brabandere, J. Emmerzaal, A. Timmermans, I. Jonkers, B. Vanwanseele, J. Davis, A Machine Learning Approach to Estimate Hip and Knee Joint Loading Using a Mobile Phone-Embedded IMU, Front. Bioeng. Biotechnol. 8 (2020). https://doi.org/10/ggsk9g.

[18] Y. Desmarais, D. Mottet, P. Slangen, P. Montesinos, A review of 3D human pose estimation algorithms for markerless motion capture, Comput. Vis. Image Underst. 212 (2021) 103275. https://doi.org/10.1016/j.cviu.2021.103275.

[19] M.A. Boswell, S.D. Uhlrich, Ł. Kidziński, K. Thomas, J.A. Kolesar, G.E. Gold, G.S. Beaupre, S.L. Delp, A neural network to predict the knee adduction moment in patients with osteoarthritis using anatomical landmarks obtainable from 2D video analysis, Osteoarthritis Cartilage. 29 (2021) 346–356. https://doi.org/10/gnt2s8.

[20] R.M. Kanko, E.K. Laende, E.M. Davis, W.S. Selbie, K.J. Deluzio, Concurrent assessment of gait kinematics using marker-based and markerless motion capture, J. Biomech. 127 (2021) 110665. https://doi.org/10.1016/j.jbiomech.2021.110665.

[21 D. Lloyd, The future of in-field sports biomechanics: wearables plus modelling compute real-time in vivo tissue loading to prevent and repair musculoskeletal injuries, Sports Biomech. 0 (2021) 1–29. https://doi.org/10/gm585p.

Acknowledgments

This work received funding from the Gesellschaft für Forschungsförderung NÖ (Research Promotion Agency of Lower Austria) within the Endowed Professorship for Applied Biomechanics and Rehabilitation Research (SP19-004).

Collapse