Does IMU redundancy improve multi-body optimization results to obtain lower-body kinematics? A preliminary study says no.

Inertial Measurement Units (IMUs) have been proposed as an ecological alternative to optoelectronic systems for obtaining human body joint kinematics. Tremendous work has been done to reduce differences between kinematics obtained with IMUs and optoelectronic systems, by improving sensor-to-segment calibration, fusion algorithms, and by using Multibody Kinematics Optimization (MKO). However, these improvements seem to reach a barrier, particularly on transverse and frontal planes. Inspired by marker-based MKO approach performed via OpenSim, this study proposes to test whether IMU redundancy with MKO could improve lower-limb kinematics obtained from IMUs. For this study, five subjects were equipped with 11 IMUs and 30 reflective markers tracked by 18 optoelectronic cameras. They then performed gait, cycling, and running actions. Four different lower-limb kinematics were computed: one kinematics based on markers after MKO, one kinematics based on IMUs without MKO, and two based on IMUs after MKO performed with OpenSense (one with, and one without, sensor redundancy). Kinematics were compared via Root Mean Square Difference and correlation coefficients to kinematics based on markers after MKO. Results showed that redundancy does not reduce differences with the kinematics based on markers after MKO on frontal and transverse planes comparatively to classic IMU MKO. Sensor redundancy does not seem to impact lower-limb kinematics on frontal and transverse planes, due to the likelihood of the "rigid component" of soft-tissue artefact impacting all sensors located on one segment.

Gait kinematics based on inertial measurement units with the sensor-to-segment calibration and multibody optimization adapted to the patient's motor capacities, a pilot study.

INTRODUCTION: Inertial Measurement Units (IMUs) offer a promising alternative to optoelectronic systems to obtain joint lower-limb kinematics during gait. However, the associated methodologies, such as sensor-to-segment (S2S) calibration and multibody optimization, have been developed mainly for, and tested on, asymptomatic subjects. RESEARCH QUESTION: This study proposes to evaluate two personalizations of the methodology used to obtain lower-body kinematics from IMUs with pathological subjects: S2S calibration and multibody optimization. METHODS: Based on previous studies, two decision trees were developed to select the best (in terms of accuracy and repeatability) S2S methods to be performed by the patient given his/her abilities. The multibody optimization was personalized by limiting the kinematic chain range of motion to the results of the subject's clinical examination. These two propositions were tested on 12 patients with various gait deficits. The patients were equipped with IMUs and reflective markers tracked by an optoelectronic system. They had to perform the postures and movements selected by the decision trees then walk back and forth along a walkway. Gait kinematics obtained from the IMUs directly (referred to as Direct kinematics), and after multibody optimization performed via the OpenSim software using the generic range of motion (referred to as Generic Optimized kinematics), and using the personalized range of motion (referred to as Personalized Optimized kinematics) were compared to those obtained with the Conventional Gait Model through Root Mean Square Errors (RMSE), Correlation Coefficients (CC) and Range of Motion differences (ΔROM). RESULTS: The RMSEs were smaller than 8.1° in the sagittal plane but greater than 7.4° in the transverse plane. The CCs, between 0.71 and 0.99 in the sagittal plane, deteriorate sharply in the frontal and transverse planes where they only measured between 0.15 and 0.68. The ΔROMs were mostly below 8.3°. Optimized kinematics did not improve compared to Direct kinematics. SIGNIFICANCE: The personalization of the proposed S2S calibration method showed encouraging results, whereas multibody optimization did not impact the resulting joint kinematics.

The contribution of multibody optimization when using inertial measurement units to compute lower-body kinematics.

Kinematics obtained using Inertial Measurement Units (IMUs) still present significant differences when compared to those obtained using optoelectronic systems. Multibody Optimization (MBO) might diminish these differences by reducing soft-tissue artefacts - probably emphasized when using IMUs - as established for optoelectronic-based kinematics. To test this hypothesis, 15 subjects were equipped with 7 IMUs and 38 reflective markers tracked by 18 optoelectronic cameras. The subjects walked, ran, cycled on an ergocycle, and performed a task which induced joint movements in the transverse and frontal planes. In addition to lower-body kinematics computed using the optoelectronical system data, three IMU-based kinematics were computed: from IMU orientations without MBO; from MBO performed using the OpenSense add-on of the OpenSim software (OpenSim 4.2, Stanford, USA); as outputs from the commercialised MVN MBO (Xsens, Netherlands). Root Mean Square Errors (RMSE), coefficients of correlations, and differences in range of motion were calculated between the three IMU-based methods and the reference kinematics. MVN MBO seems to present a slight advantage over Direct kinematics or OpenSense MBO, since it presents 34 times out of 48 (12 degrees of freedom * 4 sports activities) a mean RMSE inferior to the Direct and OpenSense kinematics. However, it was not always significant and the differences rarely exceeded 2°. This study does not therefore conclude on a significant contribution of MBO in improving lower-body kinematics obtained using IMUs. This lack of results can partly be explained by the weakness of both the kinematic constraints applied to the kinematic chain and segment stiffening. Personalization of the kinematic chain, the use of more than one IMU by segment in order to provide information redundancy, or the use of other approaches based on the Kalman Filter might increase this MBO impact.

Sensor-to-Segment Calibration Methodologies for Lower-Body Kinematic Analysis with Inertial Sensors: A Systematic Review.

Kinematic analysis is indispensable to understanding and characterizing human locomotion. Thanks to the development of inertial sensors based on microelectronics systems, human kinematic analysis in an ecological environment is made possible. An important issue in human kinematic analyses with inertial sensors is the necessity of defining the orientation of the inertial sensor coordinate system relative to its underlying segment coordinate system, which is referred to sensor-to-segment calibration. Over the last decade, we have seen an increase of proposals for this purpose. The aim of this review is to highlight the different proposals made for lower-body segments. Three different databases were screened: PubMed, Science Direct and IEEE Xplore. One reviewer performed the selection of the different studies and data extraction. Fifty-five studies were included. Four different types of calibration method could be identified in the articles: the manual, static, functional, and anatomical methods. The mathematical approach to obtain the segment axis and the calibration evaluation were extracted from the selected articles. Given the number of propositions and the diversity of references used to evaluate the methods, it is difficult today to form a conclusion about the most suitable. To conclude, comparative studies are required to validate calibration methods in different circumstances.