Abstract
Whole-body impulsive rotations were used to overcome several limitations associated with manual head impulse testing. A computer-controlled rotational chair delivered brief, whole-body, earth-vertical axis yaw impulsive rotations while eye movements were measured using video-oculography. Results from an unselected group of 20 patients with dizziness and a group of 22 control subjects indicated that the horizontal computerized rotational head impulse test (crHIT) is well-tolerated and provides an estimate of unidirectional vestibulo-ocular reflex gain comparable to results from caloric testing. This study demonstrates that the horizontal crHIT is a new assessment tool that overcomes many of the limitations of manual head impulse testing and provides a reliable laboratory-based measure of unilateral horizontal semicircular canal function.
Introduction
The head impulse test (HIT), which uses manually delivered head rotations and direct observation of eye movement has been shown to be a valuable bedside test of unilateral horizontal semicircular canal function [4]. More recently, the video head impulse test (vHIT), which also uses manually delivered head rotations but includes computerized recordings of both head and eye position, has been described [5]. The vHIT has several advantages compared with the bedside HIT including precise monitoring of head rotation and quantitative recording of eye movement, which allows detection of covert saccades and computation of unilateral VOR gain.
The purpose of this study was to determine the feasibility of implementing a new laboratory-based test of unilateral semicircular canal function, the computerized rotational head impulse test (crHIT). Using the principles of the vHIT, we developed a test that uses impulsive whole-body rotation without some of the drawbacks of the manual vHIT. The crHIT eliminates the need for a highly trained, proficient examiner, eliminates any contribution from the neck, reduces patient discomfort associated with head-on-torso rotation, and generates more data for each impulse. The obvious disadvantage of the crHIT is the necessity of a motorized rotational chair housed in a vestibular laboratory environment.
We studied an unselected group of 20 patients with dizziness to determine the feasibility of performing the crHIT in patients. Using results from the control subjects to determine normal limits, we compared crHIT results from patients with their caloric scores. Our study demonstrated that the crHIT is well-tolerated and provides information comparable to the vHIT.
Methods
This study was designed to assess the feasibility of using a clinical computer-controlled earth-vertical axis rotational chair to deliver whole-body rotational impulses comparable to the angular motion used for manual head impulse testing. We evaluated normal subjects and a convenience sample of patients with dizziness suspected of having a vestibular disorder. This study was approved by the Institutional Review Board of the University of Pittsburgh. All participants provided written informed consent before being enrolled in the study.
Subjects
Subjects included 20 unselected patients seen at a tertiary referral center for a complaint of dizziness and 22 near-aged matched control subjects. Patients included 14 females and 6 males, aged 20 to 60 years, mean 46.2±11.5 years. Each patient was evaluated clinically by an otoneurologist and underwent a battery of vestibular function tests using videonystagmography including binaural bithermal caloric testing. Caloric responses were analyzed using standard methods to compute a reduced vestibular response (RVR) score based on Jongkee’s formula. Ice water caloric testing was performed only if bithermal responses were absent in an ear. The patients were divided into three groups, as shown in Table 1, based on their caloric response. Group 1 patients (six) had unilaterally absent ice water responses in an ear, Group 2 patients had normal caloric responses (four), and Group 3 patients (ten) had a unilaterally reduced (RVR > 24%) but not absent bithermal caloric response. Control subjects included 11 females and 11 males aged 24 to 64 years, mean 38.7±14.6 years. Exclusion criteria for control subjects included a history of neurologic or otologic disease, and any abnormality on audio-vestibular testing including ocular motor testing, positional testing, caloric testing, and sinusoidal earth-vertical axis rotational testing.
Device
All testing was performed using a Neuro-Otologic Test Center supplied by Neuro Kinetics Inc. (Pittsburgh, PA). The rotational chair was computer-controlled and had a peak torque of 185 ft-lbs. Eye position was measured using binocular infrared video-oculography with a resolution of 0.1 degrees and a frame rate of 100 frames per second. The video-oculography goggles included an embedded 3-axis rate sensor to record head velocity. The rotational chair was surrounded by a light proof enclosure. Participants were secured using a four-point lap and shoulder belt system. Each participant’s head was comfortably secured using a molded foam pad that was preformed to include the occiput and cervical regions.
Procedure
Each participant was assisted into the rotational chair and comfortably and securely restrained with the head pitched forward such that Reid’s plane was in the earth-horizontal plane based on visual inspection by the audiologist who performed the testing. The head was positioned such that the axis of rotation was approximately 2 cm in front of the interaural line. The video-oculography goggles were calibrated using fixed targets at 10 degrees horizontally and vertically. Then, each subject underwent two series of rotational impulses, one consisting of nominal 750 deg/sec2 peak accelerations, the other consisting of nominal 1,000 deg/sec2 impulses. During testing, the audiologist assured that there was no obvious slippage of the video-oculography goggles by observing the location of the orbits in the video and inspecting the position of the goggles between each series of impulses. A representative series of 1,000 deg/sec2 impulses is illustrated in Fig. 1. A representative 1,000 deg/sec2 impulse is illustrated in Fig. 2. Note that each series consisted of 14 impulses, two of which were “catch trials.” For the nominal 750 deg/sec2 series, there were six rightward and six leftward impulses randomized in direction and ranging between 713 and 814 deg/sec2 peak acceleration such that no two impulses were exactly alike. These 12 impulses had an average peak acceleration of 765 + /–27.4 deg/sec2 and an average peak velocity of 150 + /–5.4 deg/sec. The two randomly placed catch trials each had a peak acceleration of 305 deg/sec2 and were not used for data analysis. Before each trial the subject was positioned directly in front of an earth-stationary visual target that remained illuminated during the rotational impulse. After each impulse, the chair was decelerated to a stop at between 150 and 200 deg/sec2. The time that the chair remained stationary prior to the subsequent impulse was random and ranged from 3 to 5 seconds.
For the nominal 1,000 deg/sec2 series, there were also six rightward and six leftward impulses randomized in direction and ranging between 945 and 1080 deg/sec2 such that no two impulses were exactly alike. These 12 impulses had an average peak acceleration of 995±41.8 deg/sec2 and an average peak velocity of 148±6.2 deg/sec. Two randomly placed catch trials had peak accelerations of 337 deg/sec2 and 405 deg/sec2 and were not used for data analysis.
Each impulse consisted of an “S” shaped velocity profile (see Fig. 2) based on the built-in LabVIEW “error function.” The function was scaled to achieve the desired peak velocity of 150 deg/sec over the amount of time necessary to achieve an average peak acceleration of 750 deg/sec2 or 1,000 deg/sec2. The peak acceleration of the movement was calculated as the slope of a best fit line through the center 25% of the calculated S-Curve. This velocity profile achieved the desired stimulus intensity while minimizing subject discomfort caused by excessive jerk, i.e., change in acceleration, associated with the onset of chair rotation.
Data analysis
The eye movement and rotational movement data were used to estimate VOR gain for each trial in each series except for the catch trials. During each impulse, data for left and right eye position and chair velocity were collected. Eye position data were inverted and eye velocity was calculated separately for the left and right eyes and then averaged. Eye velocity data were synchronized to chair velocity data using a cross-correlation procedure.
VOR gain was computed using three different methods including a velocity gain, a position gain, and an acceleration gain. The first step in the analysis was to determine if the response was valid and worthy of computing a VOR gain for that trial. Trials were excluded if a saccade occurred during the relevant portion of the stimulus (see below). Saccades were detected based upon an eye acceleration exceeding 1500 deg/sec2.
Velocity gain
Eye movement data were analyzed during that portion of the trial after chair velocity exceeded 40 deg/sec until chair velocity reached 90% of the peak velocity for that trial or the first saccade occurred. Note that trials were excluded if a saccade occurred within 40 msec of the chair velocity reaching 40 deg/sec. This region of the stimulus is illustrated in Fig. 3A for a 750 deg/sec2 trial in which no saccade occurred during the relevant portion of the stimulus. The region of the stimulus used for analysis in which a saccade occurred during the relevant portion of the stimulus is illustrated in Fig. 3B. If no saccades occurred during the relevant portion of the impulse, the total position excursion of the chair used for data analysis for eye velocity gain was 11.5 degrees for the nominal 750 deg/sec2 impulses and 15 degrees for the nominal 1,000 deg/sec2 impulses. For each data point, velocity gain was calculated using the ratio: Gain = Veye/Vchair. Then, the mean gain was calculated for each impulse. The average VOR gain for rightward impulses was calculated using all of the usable rightward impulses. The average VOR gain for leftward impulses was calculated using all of the usable leftward impulses.
Position gain
For position gain, eye movement data were analyzed from the start of the impulse to 90% of peak velocity or the occurrence of the first saccade. Note that trials were excluded if a saccade occurred less than 40 msec from the start of the impulse. Total eye excursion was calculated by mathematically integrating eye velocity. Position gain was calculated for each impulse using the following equation: Gain = (∫eye velocity)/(∫chair velocity). As for velocity gains, the position gains were averaged separately for rightward and leftward impulses.
Acceleration gain
Eye movement data were analyzed during the 100 msec interval bracketing the peak stimulus acceleration. Note that trials were excluded if a saccade occurred in this interval. For each impulse, a linear regression was calculated for eye and chair velocity data in that interval. Acceleration gain was calculated as the ratio of the slope of eye velocity to the slope of chair velocity. As for velocity and position gains, the average acceleration gain was calculated separately for rightward and leftward impulses.
Descriptive statistics were computed for each analysis method to determine the mean and standard deviation of the VOR gain values for the control subjects. We averaged 44 values for each method, reflecting each ear for the 22 control subjects. The lower limit of normal for each method was considered as the mean VOR gain minus two standard deviations.
Results
The rotational impulse test was well-tolerated by all subjects including both patients and normal controls. None of the participants asked to discontinue testing for either the nominal 750 or 1,000 deg/sec2 stimuli. Representative data from a nominal 750 deg/sec2 rotational impulse from a control subject are shown in Fig. 4. Data are illustrated during that region of the stimulus used for estimating VOR gain. As would be expected in a control subject, eye position and eye velocity match head position and head velocity very closely for both leftward and rightward impulses. VOR gain for this subject averaged across leftward and rightward stimuli was.94 using the velocity method, 0.99 using the position method, and 0.77 using the acceleration method. The means and standard deviations of the normative data are shown in Table 2 for both the nominal 750 and 1,000 deg/sec2 stimuli. Note the consistency of the data across the two stimulus intensities and note that the velocity method provided the smallest standard deviations.
Representative data from a nominal 750 deg/sec2 rotational impulse for a patient with an absent ice-water caloric response on the right is shown in Fig. 5. Data are illustrated during that region of the stimulus used for estimating VOR gain. Note that eye position and eye velocity match head position and head velocity very closely for the leftward impulse but not for the rightward impulse. VOR gain for this subject using the velocity method was 1.0 to the left and 0.57 to the right, 0.98 to the left and 0.56 to the right using the position method, and 0.97 to the left and 0.55 to the right using the acceleration method.
VOR gains for all of the patients in both directions obtained using the velocity gain method are plotted vs. total unilateral bithermal caloric response in Fig. 6A for the nominal 750 deg/sec2 impulses and in Fig. 6B for the 1,000 deg/sec2 impulses. Each data point represents either rightward VOR gain vs. caloric responsiveness of the right ear or leftward VOR gain vs. caloric responsiveness of the left ear. Note that results for nominal 750 deg/sec2 impulses and results for nominal 1,000 deg/sec2 impulses were largely the same. For ears with absent caloric responses, VOR gain was generally less than normal and for ears with a total caloric responsiveness of about 30 deg/sec or greater, VOR gain was generally in the normal range. For ears with a total caloric responsiveness less than about 30 deg/sec but not absent, results were inconsistent with respect to the lower limit of normal.
Figure 7 is a plot of VOR gain assessed with the horizontal crHIT vs. the reduced vestibular response score obtained from Jongkee’s formula for patients in all three groups. Each data point represents either rightward VOR gain or leftward VOR gain vs. that patient’s caloric weakness score. Note that results for nominal 750 deg/sec2 impulses (Fig. 7A) and results for nominal 1,000 deg/sec2 impulses (Fig. 7B) were largely the same. For the group 3 nominal 750 deg/sec2 impulses, eight of the ten patients with a clinically significant (>24%) caloric weakness had an abnormal VOR gain in the expected direction and for the nominal 1,000 deg/sec/sec impulses, seven of the ten patients with a clinically significant caloric weakness had an abnormal VOR gain in the expected direction.
Discussion
This report describes a new vestibular test, the horizontal computerized rotational head impulse test (crHIT), which uses whole-body rotation to achieve a unilateral VOR assessment comparable to the head impulse test (HIT) [4] and records eye movements with video-oculography. Current clinical rotational testing using either sinusoidal harmonic stimulation or velocity trapezoids is not considered adequate for assessing each labyrinth separately [3]. Brief whole body rotations to assess VOR initiation were described by Crane and Demer [2] using magnetic scleral search coils to record eye position. Their study showed a reduced VOR gain for rotation toward the side of lesion in four patients with surgically confirmed unilateral, peripheral vestibular deafferentiation. The bedside HIT, originally described by Halmagyi et al. [4] employed manually delivered high accelerations of the head of 2000 to 5000 deg/sec2 to silence the eighth nerve semicircular canal afferents on the contralateral side to enable a unilateral vestibular assessment using head rotation. Bedside head impulse testing has been improved by video head impulse testing (vHIT) [5], which also uses manual head impulses but uses computerized video goggles and rate sensors to measure eye and head movement to improve the accuracy of the test. Our motivation for developing the crHIT was to overcome many of the challenges associated with manual head-only impulse testing including 1) obviating the need for a highly trained proficient examiner, 2) reducing the number of impulses necessary as each impulse is precisely defined, 3) reducing patient discomfort associated with head-on-torso rotation, especially discomfort in older patients and those with limited range of motion of the neck, 4) reducing the abruptness of the movement and thus reducing unpleasant jerk, 5) allowing a random magnitude and direction of rotation thus minimizing prediction, and 6) avoiding the small position limitation of a head-on-torso rotation, which is necessarily associated with limited data for each impulse. The obvious disadvantages of the crHIT include the need for a specialized rotational chair and the inability to transport the test equipment to the bedside.
Our choice of rotational trajectory and stimulus intensity was based on a tradeoff between maximizing subject comfort and silencing or nearly silencing the contralateral semicircular canal afferents. As for the manual HIT, it is this inhibitory saturation nonlinearity of vestibular afferents that underlies the crHIT [4, 11–13]. Although crHIT uses accelerations lower than those typically used with the manual HIT, the magnitude of the impulse, which depends upon the product of acceleration and time and is thus related to peak velocity, is about the same for crHIT and manual HIT. Because subject comfort is associated with both the magnitude and duration of the acceleration and the rate of change of acceleration, i.e., jerk, we used an S-shaped velocity profile that had smooth first and second derivatives at the onset of the movement. In preliminary studies we found that accelerations that exceeded 1,000 deg/sec2 were uncomfortable. We also reasoned that the magnitude of the peak velocity should be about the same as that used for the head impulse test. Thus, we selected a peak velocity of about 150 deg/sec, which was considered adequate for a unilateral assessment technique based on the work of Halmagyi and Curthoys [4]. In this pilot study, we compared two peak acceleration profiles, one with a nominal peak acceleration of 750 deg/sec2 and the other with a nominal peak acceleration of 1,000 deg/sec2. Both of these stimulus profiles were expected to achieve predominantly unilateral excitation of the horizontal semicircular canal afferents although the stimulus profile with the lower magnitude was considered a priori to be more comfortable. We found that both of the stimulus profiles used in this study were well-tolerated with none of the subjects requesting testing be discontinued. Self-reports of discomfort during the testing were similar for both stimulus magnitudes. The standard deviation of VOR gain for the control subjects was about the same for the two acceleration magnitudes.
For this pilot study, whose aim was primarily to establish feasibility, we chose to compare our results using the crHIT with those from the clinical standard test of unilateral horizontal semicircular function, namely binaural bithermal caloric testing using warm and cool water irrigation of the external auditory canals. We recognize that caloric testing is not a gold standard and that we could not determine either sensitivity or specificity of the crHIT even with a larger sample of patients. Similar to comparisons between vHIT and caloric testing [1, 10], we found that results from the crHIT and those from caloric testing were highly correlated but not identical. Overall, VOR gain using the crHIT was usually normal if an ear had a total of warm and cool caloric scores of at least about 30 deg/sec. Also, the crHIT identified the side of caloric weakness in 70% of cases using a normal cutoff from Jongkee’s formula of 24%. These findings were about the same for both acceleration stimulus magnitudes.
In this pilot study, we also compared three data analysis methods to determine which method had the lowest variability in gain of the VOR. We found that the velocity method had the lowest variability and that the acceleration method had the most variability. The crHIT profile used in this study produced about 100 to 150 milliseconds of eye movement data for each trial suitable for assessing VOR gain. This amounts to substantially more data per impulse as compared with manual horizontal vHIT testing, which yields about 30–50 milliseconds of usable data per impulse. From the descriptions available, it appears that analyses of position [6], velocity [5], and acceleration [3] have been used for estimating VOR gain from manual vHIT testing. The lower limits of normal for VOR gain using the vHIT based on Macdougall et al. [5] is 0.68. Using the velocity method of data analysis in our study using the crHIT, the lower limit of normal for VOR gain was substantially higher than those used for the vHIT. Although the larger amount of usable data per trial and the lower variability of VOR gain imply that crHIT may be a more sensitive test than the vHIT, future studies directly comparing the crHIT with the vHIT will be required to address this issue. In this study we did not assess the time of onset or the size of saccades during the rotational impulses. Future studies of crHIT may include this potentially useful indicator of VOR compensation in patients with unilateral peripheral vestibular loss [7, 9].
Footnotes
Acknowledgments
This study was supported by NIH grant DC011884. The authors thank Anita Lieb for technical assistance.