Vestibular physical therapy improves turning not straight walking during the inertial sensor-instrumented Timed Up and Go test

1 Background: Deficits in vestibular function increase the risk for fall while turning. 2 However, the clinical assessment of turning in patients with vestibular dysfunction is 3 lacking, and evidence is limited that identifies how effective vestibular physical therapy 4 (VPT) is for improving turning performance. 5 Objective: To quantify and compare walking and turning performance during the 6 instrumented timed up and go (TUG) test using inertial measurement units (IMUs) for 7 clinical settings. We investigate novel instrumented TUG parameters for ability to 8 distinguish patients with unilateral vestibular deafferentation (UVD) from control groups, 9 and discriminate the differences in turning parameters of UVD patients following a VPT 10 program. 11 Methods: We recruited 38 patients following UVD surgery, 26 age-matched Veteran 12 patient (VA) controls with reports of non-vestibular dizziness, and 12 age-matched 13 healthy controls. Individuals were donned with body-worn IMUs and given verbal 14 instructions to complete the TUG test as fast as safely possible. The IMU-instrumented 15 and automated assessment of the TUG test provided component-based TUG 16 parameters, including the novel walking:turning ratio. Among the UVD patients, 19 17 patients completed an additional instrumented TUG testing after VPT. 18 Results: The walking:turning time ratio showed that turning performance in pre VPT 19 UVD patients are significantly more impaired than VA patients and healthy controls (p < 20 0.001). Vestibular rehabilitation significantly improved turning performance and 21 “normalized” their walking:turning time ratio compared to healthy controls (p < 0.001). However, the duration of the straight walking component was not significantly different as to that after VPT as well as healthy controls. Conclusions: Our data showed that the IMU-instrumented TUG test can distinguish 25 patients with vestibular deafferentation and objectively quantify the change in their 26 turning performance after surgery. The IMU-based instrumented TUG parameters have 27 potential to quantify the efficacy of VPT and be adopted in the clinic.


Introduction 46
Patients with vestibular hypofunction commonly present with dizziness, balance deficits, 47 7 of transitions from straight-path walking to turning while walking may provide clues as to 92 why people with vestibular hypofunction have difficulty during turning, particularly 93 relevant given the head velocities can be high [24]. Additionally, there may exist unique 94 relationships between the total time spent straight walking versus turning. In this study, 95 we quantified the walking and turning components via IMUs and introduce novel IMU-96 based TUG parameters, including the walking:turning time ratio as well as number of 97 steps ratio. The aims of this study were: 1) to examine the IMU-instrumented TUG 98 parameters during straight walking and turning in patients following unilateral vestibular 99 deafferentation surgery and two control groups; and 2) to determine the differences 100 between the IMU-instrumented TUG parameters following a progressive five-week VPT 101 program in UVD patients. We hypothesized that reduced time and number of steps to 102 perform a turn might improve after VPT.
The TUG test consists of five consecutive tasks; standing up from a chair, straight 139 walking, 180° turning, straight walking, and turning to sitting down on the chair [7]. We 140 analyzed the signal from the ankle IMUs and the chest IMU to segment in two 141 components, i.e., straight walking and turning. The chest IMU gyroscope data for yaw 142 rotation provided a turning direction (e.g. clockwise and counter-clockwise) and an 143 easily distinguished trace for the initiation of turning from a straight-path. Since the time 144 derivative of the turn-angle is the angular velocity, we calculated the trunk angular 145 displacement in the horizontal (yaw) plane by integrating the yaw angular velocity. The 146 mathematical model-based turning segmentation [18] provided the ability to 147 automatically distinguish performance of the individual tasks of the TUG (e.g., straight 148 walking and 180° turning around a cone). We applied the gait event detection algorithm 149 to the pitch angular velocity waveform recorded from both ankle IMUs. The mid-swing 150 event was selected to monitor step counts [29]. Figure 1 shows the segmented phases 151 of straight walking (yellow) and turning (green) and the detected gait event (red circle: 152 mid-swing). The end of trunk rotation was marked with a couple of steps after the 153 second straight walking component (the second yellow shaded area) and denoted the 154 end of the second turn component that occurs as participants return to seated position 155 on the chair. Both the first standing up from a chair and the second turning (returning to 156 sit on the chair) components were excluded from this below analysis.
This technique was used to determine the TUG variables, which included total time 161 All variables were normally distributed hence, parametric analysis was performed. A 175 one-way ANOVA was performed to compare variables between groups (UVD patients 176 vs VA patient controls vs healthy controls). Post hoc testing using the least significant 177 difference (LSD) correction was applied to compare between groups differences. To 178 evaluate the effect of VPT on the TUG parameters a paired t-test was performed to 179 compare the UVD patients' pre-test results vs. the post-test results. A second analysis 180 between the UVD patients' post-test results and the healthy controls was conducted by 181 independent t-test. The level of statistical significance was set at p ≤ 0.05. Mean and 182 one standard deviation (1 SD) values of the dependent variables were calculated for steps numbers; and total time and steps numbers). Descriptive statistics (mean and 1 185 SD) were used to summarize the results. 186

Results 188
Between Groups 189 patients trended to also use more steps (9%, p = 0.053) than the healthy controls. 209 During the turning component of the TUG, however, the UVD patients used a 16% 210 greater number of steps, than both the VA patient controls (p < 0.001) and the healthy 211 controls (p = 0.005). The healthy control group used fewer steps during the entire TUG 212 test compared with either UVD patients (12% less steps, p = 0.011) but not the VA 213 controls (10% less steps, p = 0.055). 214 215 Walking:turning ratio variables: The VA controls had a significantly higher ratio of 216 walking:turning steps (30% higher, p < 0.001) compared with both the UVD patients and 217 healthy controls (22% higher, p < 0.001). Only the UVD patients had a significantly 218 reduced ratio of time spent in walking:turning (p < 0.001, 19% reduction than both VA 219 and healthy controls). 220 221

Effect of Vestibular Rehabilitation 222
The UVD patients significantly improved after five weeks of VPT in most of the 223 measured TUG parameters. Figure 2 shows an example of the Pre and Post-rehab 224 results in a patient with UVD. The number of steps during the straight walking 225 component before VPT was the same as that after VPT (8 steps). During 180° turning 226 around a cone, the UVD patient 'total number of steps' reduced from five to three, 227  3.92 (s) and 3.65 (s), respectively (p = 0.06). However, after VPT, the turning time was decreased from 2.35 (s) to 1.81 (s), (p < 0.001). The walking:turning step ratio after VPT 230 was increased to 2.7 from 1.6 before VPT (p < 0.001). After VPT, the walking:turning 231 time ratio was also increased to 2.1 from 1.7 before VPT (p < 0.001).  given none of the timing variables were different than the healthy controls (0.1 < p < 261 0.361) ( Figure 3). However, the spatial properties of the UVD patients were less altered 262 given, the UVD patients used 10% more steps during the straight component of the 263 TUG compared to healthy controls (p = 0.007). During the turning component of the 264 TUG however, the UVD patients used 12% less steps than healthy controls (p = 0.011) 265 after VPT. Thus, the UVD patients walking:turning steps ratio was 28% significantly 266 higher than the healthy controls ( Figure 3). Finally, after VPT, the UVD patients' ratio of instrumented version using IMUs have been suggested and identified as having the 283 ability to classify fallers among the elderly [11,12], amputee [13], and Parkinson's 284 disease populations [17][18][19]. However, to the best of our knowledge, only one study has 285 incorporated IMUs to examine the test-retest reliability of the instrumented TUG in 286 patients with vestibular disorders and its association with fall risk [34]. The authors do 287 suggest that the instrumented TUG has potential to enable clinicians and therapists to 288 objectively assess the efficacy of their interventions in patients with vestibular disorders. 289 However, they did not use the instrumented TUG to examine the effectiveness of VPT 290 in patients with vestibular hypofunction; furthermore, no parameters have been derived 291 from IMUs to comparatively analyze the turning sub-component of the TUG in 292 association with straight walking. 293

294
Ours is the first study to distinguish patients with vestibular pathology from healthy 295 controls and non-vestibular dizzy patient controls based on the ratio between walking 296 and turning while completing the instrumented TUG. This is a critical result as it has 297 been reported that typical measures of gait, such as the 10-meter walking test, may not Our data also support this observation given the UVD patients showed no differences in 300 straight walking time before or after completing VPT. However, the UVD patients did 301 demonstrate abnormally increased walking:turning ratios due to significantly slower 302 turning, compared against two control groups. It is also interesting to note that, after 303 VPT, the UVD patients also demonstrated reduced turning time which was half of the 304 straight walking time-similar to the VA and healthy controls. Only the UVD patient 305 group before showed that the turning time was 61% of the straight walking. Our finding 306 suggests the percentage of time spent turning (50%) compared with straight-walking 307 may be a clinically meaningful parameter for distinguishing patients with vestibular 308 hypofunction as well as assessing the effectiveness of VPT in those patients. 309 310 Vestibular disorders cause changes in gait behavior [39]; however, the explicit kinematic 311 differences in turning by those with vestibular deficit remains unclear. Furthermore, 312 while there are extensive studies showing how effective VPT is for reducing dizziness 313 and falls in patients with vestibular dysfunction [5,6,10], the contributions of VPT as a 314 treatment for reducing turning difficulties are less understood. In healthy controls, 315 turning 180° around a cone during the TUG involves a smooth and continuous top-down 316 rotation from the head to the trunk [40,41] with resulting asymmetries in gait 317 parameters between limbs (i.e. stride length and stance time) [42][43]. It is presumed 318 that head movement and upper body coordination during transitions from straight-path 319 walking to turns is critical to ensure a stable position and to aid in gaze stabilization [44-orientation, gait, head movement, and upper body coordination. It was recently reported 322 that patients with unilateral vestibular hypofunction reveal fewer, smaller, and slower 323 head movements after surgery [3,4]. Additionally, these authors suggested that early 324 referral for vestibular rehabilitation may be beneficial to improve the recovery of gait,