On the use of knee functional calibration to determine the medio-lateral axis of the femur in gait analysis: Comparison with EOS biplanar radiographs as reference

How reliable are femoropelvic kinematics during deep squats? The influence of subject-specific skeletal modelling on measurement variability

BACKGROUND

Biplanar radiography displays promising results in the production of subject-specific (S.specific) biomechanical models. However, the focus has predominantly centred on methodological investigations in gait analysis. Exploring the influence of such models on the analysis of high range of motion tasks linked to hip pathologies is warranted. The aim of this study is to investigate the effect of S.Specific modelling techniques on the reliability of deep squats kinematics in comparison to generic modelling.

METHODS

8 able-bodied male participants attended 5 motion capture sessions conducted by 3 observers and performed 5 deep squats in each. Prior to each session a biplanar scan was acquired with the reflective-markers attached. Inverse kinematics of pelvis and thigh segments were calculated based on S.specific and Generic model definition. Agreement between the two models femoropelvic orientation in standing was assessed with Bland-Altman plots and paired t- tests. Inter-trial, inter-session, inter-observer variability and observer/trial difference and ratio were calculated for squat kinematic data derived from the two modelling approaches.

RESULTS

Compared to the Generic model, the S.Specific model produced a calibration trial that is significantly offset into more posterior pelvis tilt (-2.8±2.7), hip extension (-2.2±3.8), hip abduction (-1.2±3.6) and external rotation (-13.8±11.4). The S.specific model produced significantly different squat kinematics in the sagittal plane of the pelvis (entire squat cycle) and hip (between 40 % and 60 % of the squat cycle). Variability analysis indicated that the error magnitude between the two models was comparable (difference<2°). The S.specific model exhibited a lower variability in the observer/trial ratio in the sagittal pelvis and hip, the frontal hip, but showed a higher variability in the transverse hip.

SIGNIFICANCE

S.specific modelling appears to introduce a calibration offset that primarily translates into an effect in the sagittal plane kinematics. However, the clinical added value of S.specific modelling in terms of reducing experimental sources of kinematic variability was limited.

Marker location and knee joint constraint affect the reporting of overhead squat kinematics in elite youth football players

Motion capture systems are used in the analysis and interpretation of athlete movement patterns for a variety of reasons, but data integrity remains critical regardless. The extent to which marker location or constraining degrees of freedom (DOF) in the biomechanical model impacts on this integrity lacks consensus. Ten elite academy footballers performed bilateral overhead squats using a marker-based motion capture system. Kinematic data were calculated using four different marker sets with 3DOF and 6DOF configurations for the three joint rotations of the right knee. Root mean squared error differences between marker sets ranged in the sagittal plane between 1.02 and 4.19 degrees to larger values in the frontal (1.30-6.39 degrees) and transverse planes (1.33 and 7.97 degrees). The cross-correlation function of the knee kinematic time series for all eight marker-sets ranged from excellent for sagittal plane motion (>0.99) but reduced for both coronal and transverse planes (<0.9). Two-way ANOVA repeated measures calculated at peak knee flexion revealed significant differences between marker sets for frontal and transverse planes (p < 0.05). Pairwise comparisons showed significant differences between some marker sets. Marker location and constraining DOF while measuring relatively large ranges of motion in this population are important considerations for data integrity.

On the accuracy of the Conventional gait Model: Distinction between marker misplacement and soft tissue artefact errors

There is a lack of knowledge about the accuracy of the Conventional Gait Model (CGM), compared to the true bone motion. Accuracy is hindered by both marker misplacement and soft-tissue artefact (STA). The effect of the lateral knee marker (KNE) misplacement and STA was determined from a secondary analysis of 13 subjects equipped with a total knee prothesis for which simultaneous dual-plane fluoroscopy and marker-based motion capture was available. In average, STA alone led to 3.3°, 2.9° and 6.7° errors for knee flexion, knee abduction, and the absolute hip rotation respectively. In comparison, marker misplacement led to 0.9°, 4.0° and 12.3° errors for the same kinematics. We showed that STA alone may lead to knee flexion-adduction cross-talk. This finding has clinical repercussions for the use of knee cross talk as a qualitative indicator of knee axis alignment. Our study showed that cumulative effects of marker misplacement and STA affect the transverse plane angles, making challenging to track internal/external rotation with less than 5° of errors.

The effect of functional calibration methods on gait kinematics in adolescents with idiopathic rotational deformity of the femur

BACKGROUND

Due to anatomical deviations, assumptions of the conventional calibration method for gait analysis may be violated in individuals with rotational deformities of the femur. Functional calibration methods were compared with conventional methods in this group for 1) localization of the hip joint center and orientation of the knee axis, and 2) gait kinematics.

METHODS

Twenty-four adolescents with idiopathic rotational deformity of the femur underwent gait analysis and a CT scan. During standing, distance between hip joint centers and knee axis orientation were compared between calibration methods, with CT serving as reference for hip joint center estimation. Gait kinematics were compared using statistical parametric mapping.

FINDINGS

The conventional calibration method estimated the hip joint center closer to the CT reference (4±12 mm more lateral) than the functional calibration method (26 ± 20 mm more lateral). Orientation of the knee joint axis was 2.6° less internal in the functional calibration method. During gait, statistical parametric mapping revealed significantly more hip flexion, less external hip rotation during the swing phase, less knee varus-valgus motion, and larger knee flexion angles when applying the functional method.

INTERPRETATION

Functional calibration methods were less accurate in determining the hip joint center location than the conventional calibration method and resulted in a knee joint axis that was less internally rotated. Importantly, there was less knee joint angle crosstalk during gait when using the functional method. Although differences between methods on gait kinematics were within clinically acceptable limits for the sagittal plane, relatively larger differences on transversal hip kinematics may hold clinical importance.

Are we overestimating internal rotation gait by conventional modelling?

BACKGROUND

The determination of the knee joint axis (KJA) via specific calibration movements has become a promising alternative to the conventional approach to determine this axis based on regression equations or directly via marker placement on bony landmarks of the knee. Since the orientation of the KJA may differ between methods, it has direct influence on hip rotation and may therefore influence clinical decision making in context of transverse plane gait deviations.

RESEARCH QUESTION

Does a functional KJA calibration lead to clinically relevant differences in hip rotation estimates during gait compared to the conventional method?

METHODS

95 subjects (age: 19.9 years; BMI: 21.1 kg/m²), including 71 patients with potential rotation malalignment, were prospectively examined and underwent 3D gait analysis. For the conventional approach the KJA was determined by applying a knee alignment device, for the functional approach subjects were asked to perform two different calibration movements. Each procedure was performed twice. Mean hip rotation in stance (mHipRotSt) was determined following the conventional and the functional KJA calibration.

RESULTS

Deming regression analysis for the comparison of conventional and functionally measured hip rotation revealed a systematic and substantial difference between methods (slope: 0.63; intercept: 0.31°). Measurement repetition with the knee alignment device revealed typical errors around 5°, whereas the functional methods lead to profoundly smaller errors around 1-2° with slightly inferior results for SQUAT compared to FLEX. However, when compared to conventional frontal plane video-taping, the conventional method seemed to reflect the more consistent results.

SIGNIFICANCE

The systematic linear discrepancy in mHipRotSt obtained by a functional approach as compared to the conventional approach appears critical since patients with severe internal or external rotation gait may be misjudged in function when receiving corrective procedures such as femoral derotation osteotomy.

Measurement error associated with gait cycle selection in treadmill running at various speeds

A common approach in the biomechanical analysis of running technique is to average data from several gait cycles to compute a 'representative mean.' However, the impact of the quantity and selection of gait cycles on biomechanical measures is not well understood. We examined the effects of gait cycle selection on kinematic data by: (i) comparing representative means calculated from varying numbers of gait cycles to 'global' means from the entire capture period; and (ii) comparing representative means from varying numbers of gait cycles sampled from different parts of the capture period. We used a public dataset (n = 28) of lower limb kinematics captured during a 30-second period of treadmill running at three speeds (2.5 m s^-1, 3.5 m s^-1 and 4.5 m s^-1). 'Ground truth' values were determined by averaging data across all collected strides and compared to representative means calculated from random samples (1,000 samples) of n (range = 5-30) consecutive gait cycles. We also compared representative means calculated from n (range = 5-15) consecutive gait cycles randomly sampled (1,000 samples) from within the same data capture period. The mean, variance and range of the absolute error of the representative mean compared to the 'ground truth' mean progressively reduced across all speeds as the number of gait cycles used increased. Similar magnitudes of 'error' were observed between the 2.5 m s^-1 and 3.5 m s^-1 speeds at comparable gait cycle numbers -where the maximum errors were < 1.5 degrees even with a small number of gait cycles (i.e., 5-10). At the 4.5 m s^-1 speed, maximum errors typically exceeded 2-4 degrees when a lower number of gait cycles were used. Subsequently, a higher number of gait cycles (i.e., 25-30) was required to achieve low errors (i.e., 1-2 degrees) at the 4.5 m s^-1 speed. The mean, variance and range of absolute error of representative means calculated from different parts of the capture period was consistent irrespective of the number of gait cycles used. The error between representative means was low (i.e., < 1.5 degrees) and consistent across the different number of gait cycles at the 2.5 m s^-1 and 3.5 m s^-1 speeds, and consistent but larger (i.e., up to 2-4 degrees) at the 4.5 m s^-1 speed. Our findings suggest that selecting as many gait cycles as possible from a treadmill running bout will minimise potential 'error.' Analysing a small sample (i.e., 5-10 cycles) will typically result in minimal 'error' (i.e., < 2 degrees), particularly at lower speeds (i.e., 2.5 m s^-1 and 3.5 m s^-1). Researchers and clinicians should consider the balance between practicalities of collecting and analysing a smaller number of gait cycles against the potential 'error' when determining their methodological approach. Irrespective of the number of gait cycles used, we recommend that the potential 'error' introduced by the choice of gait cycle number be considered when interpreting the magnitude of effects in treadmill-based running studies.

Which typical floor movements of men's artistic gymnastics result in the most extreme lumbar lordosis and ground reaction forces?

Back pain is prevalent among gymnast populations and extreme flexion or extension of the lumbar spine along with high ground reaction forces (GRFs) are known to increase intervertebral stress. The aim of this study was to determine which postures and dynamic conditions among common floor movements provide the greatest risk of injury in men's artistic gymnastics (MAG). For this purpose, lumbar spine curvatures, obtained through a full-body subject-specific kinematic model fed by motion capture data, and GRFs on feet and hands were compared between typical floor movements of MAG (pike jump, round off back handspring, front handspring, forward and backward tucked somersaults) performed by six adolescent gymnasts. The round off back handspring and the pike jump resulted respectively in the largest lumbar extension and flexion, and the forward tucked somersault take-off in the highest GRF. At ground impacts, the largest lumbar flexion was during the backward tucked somersault landing and only the back handspring hands ground contact phase led to lumbar extension. Such identification of high-risk conditions should enable better back pain management in gymnastics through more tailored training adaptations, particularly in case of pathologies or musculoskeletal specificities.

A novel dataset and deep learning-based approach for marker-less motion capture during gait

BACKGROUND

The deep learning-based human pose estimation methods, which can estimate joint centers position, have achieved promising results on the publicly available human pose datasets (e.g., Human3.6 M). However, these datasets may be less efficient for gait study, particularly for clinical applications, because of the limited number of subjects, their homogeneity (all asymptomatic adults), and the errors introduced by marker placement on subjects' regular clothing.

RESEARCH QUESTION

How a new human pose dataset, adapted for gait study, could contribute to the advancement and evaluation of marker-less motion capture systems?

METHODS

A marker-less system, based on deep learning-based pose estimation methods, was proposed. A new dataset (ENSAM dataset) was collected. Twenty-two asymptomatic adults, one adult with scoliosis, one adult with spondylolisthesis, and seven children with bone disease performed ten walking trials, while being recorded both by the proposed marker-less system and a reference system - combining a marker-based motion capture system and a medical imaging system (EOS). The dataset was split into training and test sets. The pose estimation method, already trained on the Human3.6 M dataset, was evaluated on the ENSAM test set, then reevaluated after further training on the ENSAM training set. The joints coordinates were evaluated, using Bland-Altman bias and 95 % confidence interval, and joint position error (the Euclidean distance between the estimated joint centers and the corresponding reference values).

RESULTS

The Bland-Altman 95 % confidence intervals were substantially improved after finetuning the pose estimation method on the ENSAM training set (e.g., from 106.9 mm to 17.4 mm for the hip joint). With the new dataset and approach, the mean joint position error varied from 6.2 mm for ankles to 21.1 mm for shoulders.

SIGNIFICANCE

The proposed marker-less system achieved promising results in terms of joint position errors. Future studies are necessary to assess the system in terms of gait parameters.

Development and evaluation of a new methodology for Soft Tissue Artifact compensation in the lower limb

Skin Marker (SM) based motion capture is the most widespread technique used for motion analysis. Yet, the accuracy is often hindered by Soft Tissue Artifact (STA). This is a major issue in clinical gait analysis where kinematic results are used for decision-making. It also has a considerable influence on the results of rigid body and Finite Element (FE) musculoskeletal models that rely on SM-based kinematics to estimate muscle, contact and ligament forces. Current techniques devised to compensate for STA, in particular multi-body optimization methods, often consider simplified joint models. Although joint personalization with anatomical constraints has improved kinematic estimation, these models yet don't represent a fully reliable solution to the STA problem, thus allowing us to envisage an alternative approach. In this perspective, we propose to develop a conceptual FE-based model of the lower limb for STA compensation and evaluate it for 66 healthy subjects under level walking motor task. Both hip and knee joint kinematics were analyzed, considering both rotational and translational joint motion. Results showed that STA caused underestimation of the hip joint kinematics (up to 2.2°) for all rotational DoF, and overestimation of knee joint kinematics (up to 12°) except in flexion/extension. Joint kinematics, in particular the knee joint, appeared to be sensitive to soft tissue stiffness parameters (rotational and translational mean difference up to 1.5° and 3.4 mm). Analysis of the results using alternative joint representations highlighted the versatility of the proposed modeling approach. This work paves the way for using personalized models to compensate for STA in healthy subjects and different activities.

Focusing on functional knee parameter determination to develop a better clinical gait analysis protocol

BACKGROUND

Attempts to improve protocol standards of marker-based clinical gait analysis (CGA) have been one of the main focuses of research to enhance robustness and reliability outcomes since the 1990s. Determining joint centres and axes constitutes an important aspect of those protocols. Although the hip joint is more prominent in such studies, knee joint center (KJC) and axis (KJA) directly affect all outcomes.

RESEARCH QUESTION

What recommendations arise from the study of the scientific literature for determining knee joint parameters (KJP) for protocols of CGA?

METHODS

A systematic, electronic search was conducted on November 2018 using three databases with the keyword combination ("functional approach" OR "functional method" OR "functional calibration") AND ("hip joint" OR "knee joint" OR "ankle joint") and analyzed by four reviewers. Given the existence of a recent review about the hip joint and the lack of material about the ankle joint, only papers about the knee joint were kept. The references cited in the selected papers were also screened in the final round of the search for these publications. The quality of the selected papers was assessed and aspects regarding accuracy, repeatability, and feasibility were thoroughly considered to allow for a comparison between studies. Technical aspects, such as marker set choice, KJP determination techniques, demographics, and functional movements, were also included.

RESULTS

Thirty-one papers were included and on average received a rating of about 75 % according to the quality scale used. The results showed that functional methods are superior or equivalent to predictive methods to estimate the KJA, while a regression method was slightly better for KJC prediction.

SIGNIFICANCE

Calibration methods should be applied to CGA whenever feasibility is reached. No study to date has focused on evaluating the in vivo RoM required to obtain reliable and repeatable results and future work should aim in this direction.

Coordinate system requirements to determine motions of the tibiofemoral joint free from kinematic crosstalk errors

The relative rigid body motions between the femur and the tibia (termed tibiofemoral kinematics) during flexion activities can provide an objective measure of knee function. Clinically meaningful tibiofemoral kinematics are defined as the six relative rigid body motions expressed in a joint coordinate system where the motions about and along the axes conform to clinical definitions and are free from kinematic crosstalk errors. To obtain clinically meaningful tibiofemoral kinematics, coordinate systems must meet certain requirements which neither have been explicitly stated nor in fact satisfied in any previous publication known to the author. Starting with the joint coordinate system of Grood and Suntay (1983) where motions conform to clinical definitions, the body-fixed axes must correspond to the functional (i.e. actual) axes in flexion-extension and internal-external axial rotation to avoid kinematic crosstalk errors in rotations and both functional axes must be body-fixed throughout knee flexion. To avoid kinematic crosstalk errors in translations, the origins of the femoral and tibial Cartesian coordinate systems, which serve as stepping stones for computing translations, must lie on the functional body-fixed axes. Neither the paper by Grood and Suntay nor the ISB recommendation (Wu et al., 2002) which adopted the joint coordinate system of Grood and Suntay explains these requirements. Indeed meeting these requirements conflicts with the ISB recommendation thus indicating the need for revision to this recommendation. Future studies where clinically meaningful tibiofemoral kinematics are of interest should be guided by the requirements described herein.

Accuracy and kinematics consistency of marker-based scaling approaches on a lower limb model: a comparative study with imagery data

Medical images are not typically included in protocol of motion laboratories. Thus, accurate scaling of musculoskeletal models from optoelectronic data are important for any biomechanical analysis. The aim of the current study was to identify a scaling method based on optoelectronic data, inspired from literature, which could offer the best trade-off between accurate geometrical parameters (segment lengths, orientation of joint axes, marker coordinates) and consistent inverse kinematics outputs (kinematic error, joint angles). The methods were applied on 26 subjects and assessed with medical imagery building EOS-based models, considered as a reference. The main contribution of this paper is to show that the marker-based scaling followed by an optimisation of orientation joint axes and markers local coordinates, gives the most consistent scaling and joint angles with EOS-based models. Thus, when a non-invasive mean with an optoelectronic system is considered, a marker-based scaling is preliminary needed to get accurate segment lengths and to optimise joint axes and marker local coordinates to reduce kinematic errors.AbbrevationsAJCAnkle joint centreCKEcumulative kinematic errorDoFdegree of freedomEBEOS-basedHBheight-basedHJChip joint centreKJCknee joint centreMBmarker-basedMSMmusculoskeletal modelsSPMstatistical parametric mappingSTAsoft tissue artifactEB_a.m∗EOS-based with optimised joint axes, and all model markers coordinatesMB_a.m∗marker-based with optimised joint axes, and all model markers coordinatesMB_l.a.mmarker-based with optimised segment lengths, joint axes, and selected model markers coordinatesASISanterior superior illiac spinePSISposterior superior illiac spine.

Correcting lower limb segment axis misalignment in gait analysis: A simple geometrical method

BACKGROUND

Obtaining precise and repeatable measurements is essential to clinical gait analysis. However, defining the thigh medial-lateral axis segment remains a challenge, with particular implications for the hip rotation profile. Thigh medial-lateral axis misalignment modifies the hip rotation profile and can result in a phenomenon called crosstalk, which increases knee adduction-abduction amplitude artificially.

RESEARCH QUESTION

This study proposes an a posteriori geometrical method based solely on segment anatomy that aims to correct the thigh medial-lateral axis definition and crosstalk-related error.

METHODS

The proposed method considers the thigh medial-lateral axis as the normal to the mean sagittal plane of the lower limb defined by hip, knee and ankle joint centres during one gait cycle. Its performance was compared to that of an optimisation method which repositions the axis to reduce knee abduction-adduction variance. An existing dataset was used: 75 patients with a knee prosthesis undergoing gait analysis three months and one-year post-surgery. Three-dimensional hip and knee angles were computed for two gait analysis sessions. Crosstalk was quantified using both the coefficient of determination (r²) between knee flexion-extension and adduction-abduction and the amplitude of knee adduction-abduction. The reproducibility of hip internal-external rotation was also quantified using the inter-trial, inter-session and inter-subject standard deviations and the intraclass coefficient (ICC).

RESULTS

Crosstalk was significantly reduced from r² = 0.67 to r² = 0.51 by the geometrical method but remained significantly higher than with the optimisation method with a r² < 0.01.

SIGNIFICANCES

Both methods allowed to improve the hip internal-external reproducibility from poor to moderate (original data: ICC = 0.34, geometrical method: ICC = 0.65, optimisation method ICC = 0.73). One advantage of the geometrical method is that, unlike the optimisation method, it does not require much movement, making it suitable for a wider range of patients.

Subject specific finite element mesh generation of the pelvis from biplanar x-ray images: application to 120 clinical cases

Several Finite Element (FE) models of the pelvis have been developed to comprehensively assess the onset of pathologies and for clinical and industrial applications. However, because of the difficulties associated with the creation of subject-specific FE mesh from CT scan and MR images, most of the existing models rely on the data of one given individual. Moreover, although several fast and robust methods have been developed for automatically generating tetrahedral meshes of arbitrary geometries, hexahedral meshes are still preferred today because of their distinct advantages but their generation remains an open challenge. Recently, approaches have been proposed for fast 3D reconstruction of bones based on X-ray imaging. In this study, we adapted such an approach for the fast and automatic generation of all-hexahedral subject-specific FE models of the pelvis based on the elastic registration of a generic mesh to the subject-specific target in conjunction with element regularity and quality correction. The technique was successfully tested on a database of 120 3D reconstructions of pelvises from biplanar X-ray images. For each patient, a full hexahedral subject-specific FE mesh was generated with an accurate surface representation.

Hip- and patellofemoral-joint loading during gait are increased in children with idiopathic torsional deformities

BACKGROUND

Torsional deformities of the femur and tibia are associated with gait impairments and joint pain. Several studies have investigated these gait deviations in children with cerebral palsy. However, relatively little is known about gait deviations in children with idiopathic torsion and debate ensues about the management of these patients.

RESEARCH QUESTION

What are the effects of idiopathic increased femoral neck anteversion and external tibial torsion on lower-limb kinematics, kinetics and joint loading during gait in children and adolescents.

METHODS

Patient-specific musculoskeletal models were created for 12 children/adolescents (mean age of 14 years) with torsional deformities using low-dose biplane radiographic imaging and 3D gait analysis. Comparisons of joint motion and net joint torques during gait were made to an age-matched control group with no torsional deformities. The effects of torsional deformities on muscle and joint contact forces were investigated using two personalised musculoskeletal models: one with normal torsion and another with patient-specific torsion.

RESULTS

Femoral neck anteversion and external tibial torsion for the patients were (mean ± SD) 38° ± 9° and 40° ± 10°, respectively. Patients had increased internal hip rotation and external knee rotation as well as increased pelvic tilt during gait. Additionally, the efficacy of the plantarflexor-knee extension mechanism was diminished. Hip joint contact force was higher in the model with patient-specific torsion. The mediolateral component of the patellofemoral joint contact force was also increased despite the magnitude of the resultant patellofemoral contact force being unchanged.

SIGNIFICANCE

It has been previously established that idiopathic lower-limb torsional deformities alter gait kinematics. However, this study also showed that loading of the hip and patellofemoral joints are increased. This is an important insight for the clinical management of these patients and highlights that idiopathic lower-limb torsional deformities are not a purely cosmetic issue.

Computation of hip rotation kinematics retrospectively using functional knee calibration during gait

BACKGROUND

Hip rotation kinematics during gait is a key parameter to support clinical decision making, for example in children with lower limb torsional deformities. However, hip rotation kinematics is also one of the least repeatable parameter because it is difficult to locate the position of the medio-lateral axis of the femur. Functional knee calibration provides an alternative to locate the medio-lateral axis of the femur and may be performed retrospectively, using the movement of the knee joint during gait. Although not necessarily more anatomically accurate, functional calibration may lead to increased repeatability between sessions, which would be useful to compare gait analysis data from sessions pre- and post-treatment, or to reprocess data in large gait databases.

METHODS

This study presents a workflow to perform knee functional calibration using knee kinematics during gait and update hip rotation kinematics accordingly. The workflow was applied to investigate the inter-subject, inter-session and inter-trial variance components of multiple calibration methods in a group a 10 typically developing children.

RESULTS

Results indicated that one or two degrees of freedom functional calibration methods were more repeatable inter-session (SD: 1.8°) than conventional calibration using the knee alignment device (SD: 4.7°). However, simulated reduced range of movement at the knee during gait increased inter-session variance for the functional calibration algorithms. Functional calibration did not provide any improvement over the conventional calibration when knee range of movement was reduced and flexion greater than 20° during gait, i.e. 'crouch gait'.

SIGNIFICANCE

The workflow presented allows the re-processing of gait analysis data using knee kinematics during gait only. The workflow may also be used to investigate functional axes of other joints, for example the ankle.