IMU Calibration Effect on Lower Limbs Kinematics Against Optical Motion Capture in Post-Stroke Gait

Abstract

Background

Stroke is the most common cause of disabilities worldwide. Rehabilitation is central to restore functions. Inertial measurement units (IMU) can be used to ease goal settings and monitor progression. Contrary to optical motion capture (OMC), IMU are less expensive, portable, and allow large scale data collections in ambulatory settings. Although Xsens MVN system validity has been demonstrated in healthy participants, its validity among post-stroke (PS) patients is yet to be proven.

Research question

Computation methods being affected by the calibration type; the goal of this study is to compare lower limbs kinematics from Xsens system, after two calibrations against OMC in slow PS walkers exhibiting reduced ranges of movements.

Methods

Data was collected for six PS patients. They were equipped with 29 reflective markers and seven IMU. A minimum of two walks with a dynamic calibration and four walks with a static calibration were performed. All trials were accomplished at a self-selected walking speed and PS used their usual walking aids.

Results

Few interactions between the calibration type and side were found for the ankle abduction/adduction (A/A) bias, root mean square error (RMSE), and range of motion difference (ROMd) (p = 0.011, p = 0.048, p = 0.039). Few effects of the side on errors' values were found. We noticed some effects of the calibration type on errors' values, the dynamic calibration showing better results. In the sagittal plane, we reported RMSE values from 3.6 to 4.8°, 5.2 to 6.5°, and 5.0 to 5.9° for the hip, knee, and ankle dynamic calibration.

Significance

The calibration type, reduced range of movement, and slow walking speed does not seem to impact Xsens' accuracy to a great extent. Nevertheless, dynamic calibration provides slightly better results. Considering the patient's walking ability, we recommend using this calibration.