**4. Discussion**

In this study we investigated the validity of IMUs to detect and measure head and trunk ROM and peak rotational velocity during a set of commonly prescribed vestibular rehabilitation tasks. Our findings suggest excellent validity for the IMU system when capturing head movements in both the L/R and U/D conditions and excellent validity capturing trunk movements in the L/R conditions. Inertial sensors showed moderate to excellent ability to estimate trunk ROM in the U/D conditions and was moderate to good at capturing peak rotational velocity in U/D conditions; except during tandem walking which showed poor agreement. The excellent agreement found here for head motion is consistent with previous findings [12], who showed excellent agreement for cervical angles collected using inertial sensors on the head and neck. This work also extends prior research, by identifying that IMUs can accurately estimate the ROM and peak turning velocity during both standing and locomotor tasks.

Automatically characterizing head and trunk movements during routinely prescribed vestibular exercises using IMUs is an innovative approach that will allow a more sensitive and objective analysis of progression during vestibular rehabilitation. In people with mTBI, smaller and slower head movements during performance tasks have been reported [5] but such movements are not easily quantified with the naked eye and may not be perceived by the patient performing the exercise. Quantifying such information with IMUs could inform both the treating physical therapist and, with time, the patient themselves by providing immediate feedback on velocity and quality of performance.

Despite the good agreement between IMU and motion capture systems, we believe some of the estimation errors might be attributed to a misalignment of the IMU frame relative to the anatomical axes of rotation. When the IMUs are attached to different body segments, they are not perfectly aligned with the segments' main axes of rotation. To estimate this misalignment, we asked the study participants to remain stationary in a neutral pose for about three seconds at the beginning of the recording. This information was then used to realign the sensors' data for analysis, using matrix rotation, before calculating the joint metrics. While this addressed the misalignment of the sensors relative to the anatomical axes, it assumed the participant could both remain stationary and adopt a truly neutral initial pose. This was not always true for every participant, and we hypothesize that this contributed to the larger errors observed in a few of the subjects.

Although 3D motion capture is commonly classified as a gold-standard measurement, it is possible that reduced agreement in some cases could partially be a function of the motion capture methods implemented in this study. Firstly, the accuracy of motion capture systems can decrease as the capture volume increases [21]. Despite using a 12-camera system it is possible that the size of our capture area, elongated to collect the tandem and walking trials, played a role in the reduced agreement between the two systems. Another source of disagreement between the inertial and optical systems could be attributed to skin motion artifact and muscle movements—a known issue [22] with systems that use markers attached to the body. Similarly, skin artifacts can also influence the inertial sensor measurements resulting in potential orientation changes. These orientation changes may produce joint metric estimates that are biomechanically unlikely and lead to a disagreement between the systems.
