Assessing Misperception of Rotation in Benign Paroxysmal Positional Vertigo with Static and Dynamic Visual Images

Abstract

Background:

Perception of self-motion is difficult for patients to describe. In addition, the relationship between perceived rotation and eye movements is poorly understood, because most studies of patients have investigated only static orientation.

Objective:

First, to determine whether patients with benign paroxysmal positional vertigo (BPPV) can use visual images to report perceived rotation elicited by the Dix-Hallpike maneuver. Second, to determine if the direction of patients’ perceptions align with data on classical nystagmus direction. Methods: After the Dix-Hallpike maneuver, BPPV patients viewed images -- sketches or video animations -- representing possible perceived motions. They selected one or more images representing perception.

Results:

All subjects could select images. The directions of the videos were most often backward pitch and/ or ipsilateral roll and yaw relative to body orientation in the supine Dix-Hallpike position, generally consistent with the canal stimulus. Perceived direction of rotation was statistically significantly different from the direction of eye movements as published previously, suggesting a difference in mechanisms for perception and eye movements.

Conclusion:

Patients can easily learn to use a video language to describe their experiences. Perception is generally aligned with canal stimulus and nystagmus, but not exactly.

1. Introduction

Patients with vestibular disorders complain of vertigo and other forms of spatial disorientation. Relatively few studies, however, have examined patients’ perceptions. Perception is difficult to describe in words. A better method is to ask subjects to view a visual stimulus that reflects their perception, such as a tilted line that can be turned [4, 6, 13, 30]. In patients with benign paroxysmal positional vertigo (BPPV) subjective visual vertical is impaired under some test conditions [11, 16, 20]. Anecdotally, BPPV patients have also reported vertigo in the form of illusory rotation during routine activities of daily living that typically involve pitch rotations of the head or otherwise stimulate the posterior semicircular canals, and also immediately after positional testing. No experimental studies have examined dynamic motion perception in BPPV patients, however.

Perceived motion may parallel eye movements. For example, in the context of the perceived vertical, several studies on static orientation have shown a relationship between perceived vertical and static ocular counter-roll [19, 40]. The relationship between more dynamic perception and reflexive eye movements, however, remains unclear. In a study of healthy subjects, differences between perception and reflexive eye movement directions were found during complex motion with changing rotational axes [22, 28]. Perception and eye movements appeared to match in unidimensional directions, such as left-right and up-down, in general, but were not aligned when considered in three dimensions. Other studies have shown a mismatch in timing between the perceived motion and the reflexive eye movements during periodic motion even though the directions matched [29, 42].

In a study of healthy subjects given caloric stimulation, the position of the subject made a difference, affecting perception of rotation but not reflexive eye movements [25]. When in a seated position, subjects reported significantly more perception of yaw and pitch but less roll rotation than when in a supine position. In fact, no supine subjects reported perceived pitch rotation. The eye movement directions were the same regardless of test position.

The pathognomonic pattern of nystagmus in response to the Dix-Hallpike maneuver in BPPV patients [7] provides a useful model for studying the directions of nystagmus and motion perception in a patient population. In the Dix-Hallpike maneuver, the clinician assists the patient to lie supine with the head turned approximately 45° toward the test side and with the head pitched backward approximately 30°. The test is positive if nystagmus is observed. In classical nystagmus, after a delay of several seconds following the stimulus, nystagmus occurs with the quick phase beating upward as well as ipsilaterally in roll and yaw [8, 14].

The primary goal of the present study was to determine whether BPPV patients could use a visual method to report dynamic perception, with static images or video animations that may mirror their perceived motions. A secondary goal was to determine if the direction of perceived rotation aligns with that of classical nystagmus based upon previously published data.

2. Methods

2.1. Subjects

Subjects were patients referred by board-certified otolaryngologists and neurologists for treatment of unilateral posterior canal BPPV. Diagnoses were based on clinical examinations, unilateral positive responses to Dix-Hallpike maneuvers, and -- if ordered by the physician in order to rule out other disorders -- a standard battery of other objective diagnostic tests of the vestibular system including electronystagmography (ENG), bi-thermal caloric tests, low frequency sinusoidal rotations in darkness, and cervical vestibular evoked myogenic potentials. No patients had taken vestibular-suppressant medication for at least 24 hours. All subjects were free of neurologic conditions and significant musculoskeletal disorders and had cervical range of motion within functional limits. The age range, predominance of women and predominance of right side BPPV were all consistent with reports in the literature [3, 10, 39]. See Table 1. To view the sketches and videos subjects wore their corrective lenses if needed.

Table 1.

Demographic details of the sample

Experiment	Mean age (SD, range)	Sex	N per group	Number of subjects per side of BPPV
Static 1	60.8 (13.2, 34–83)	17 female 3 male	20	11 right 9 left
Static 2	67.6 (14.1, 40–90)	14 female 6 male	20	13 right 7 left
Video 1	61.12 (13, 37–83)	21 female 4 male	25	15 right 10 left
Video 2	68 (11.5, 44–86)	16 female 4 male	20	14 right 6 left

A total of 85 subjects were recruited. New subjects were recruited for each experiment. Each subject participated in only one experiment. All subjects gave written informed consent prior to participation. This study was approved by the Institutional Review Board for Baylor College of Medicine and Affiliated Hospitals.

2.2. Instrumentation

Subjects viewed static sketches (Figs. 1 and 2) or video animations (Fig. 3) after the Dix-Hallpike maneuver, as explained in 2.3 Testing methods. The animations were developed using the MegaPOV version of POV-Ray (Persistence of Vision (TM) Raytracer, Persistence of Vision Pty. Ltd., Williamstown, Victoria, Australia) to create the frames. Six cartoon videos were developed for the Dix-Hallpike maneuver to the right, and six to the left. In each animation, the cartoon person was lying supine in the Dix-Hallpike position with its head off the treatment table. Movement of stripes on the head represented perception of rotation about a particular axis. The axes were the earth-cardinal axes, i.e. body-referenced pitch, roll and yaw, and the directions in the six animations were backward pitch, forward pitch, rightward roll, leftward roll, rightward yaw, and leftward yaw. In the animations, the head was oriented at 45° in yaw (either right or left) plus 20° pitch back, so the body-referenced pitch animations showed rotation in the posterior canal planes. The same stripe rotation speed of 45°/second was used in all animations.

Figure 1.

Venn diagram showing numbers of subjects selecting each combination of images in Picture Experiment 1. Each of the three circles represents perceived motion in one plane, as indicated by the images. The numbers in each region of overlapping circles indicates the number of subjects who selected two or more perceived motions. Numbers in regions within only one circle indicate the number of subjects that perceived only one motion. Horizontal motion was never selected, and vertical motion was selected only once and by itself.

Figure 2.

Venn diagram showing numbers of subjects selecting each combination of images in Picture Experiment 2. The larger numbers indicate subjects indicating the typical direction, whereas the smaller, “opp.”, numbers indicate subjects indicating opposite the typical direction; one subject indicated that roll was contralateral. Only one diagonal motion was selected, by itself.

Figure 3.

Venn diagram showing numbers of subjects selecting each combination of videos in Video Experiment 1. The larger numbers indicate subjects selecting videos with stripes rotating in the typical direction, which this figure shows with arrows, whereas the smaller, “opp.”, numbers indicate subjects whose choices included at least one video with oppositely-directed stripes. The actual videos had rotating stripes and no arrow; an example full video frame is shown at upper right.

2.3. Testing methods

In all experiments the corresponding author tested subjects with the Dix-Hallpike maneuver on the side of the diagnosed impairment. After subjects sat up they were shown either five sketches (Picture Experiments (PE) 1 and 2) or six videos (Video Experiments (VE) 1 and 2), with differences between the experiments explained below. In all cases, subjects were asked to indicate which sketches or videos best illustrated the subjective direction of perceived motion while in the stationary supine stage of the Dix-Hallpike maneuver. They were allowed to take as much time as needed and to view the images and videos as many times as needed. Following the convention of Guedry et al. in their Experiments 1 and 2 [22], subjects provided data after they experienced the stimulus. (Asking subjects to view the sketches or videos during the Dix-Hallpike maneuver -– with their heads upside down and while they were experiencing vertigo -- was not practical.) The investigator explained to subjects that each image or video represented up/down, back-and-forth, and tilting motions of the head.

In PE 1 and 2 subjects could keep their eyes open during the Dix-Hallpike maneuver if desired. In PE 1 subjects viewed sketches illustrating pitch, roll and yaw head rotations and vertical and left-right horizontal translations (Fig. 1). In PE 2 subjects viewed the three pitch, roll and yaw images from PE 1 and two diagonal motions: bending downward toward the right and upward toward the left, and bending downward toward the left and upward toward the right (Fig. 2). In PE 2 subjects also indicated whether the perceived motion was backward or forward (pitch), leftward or rightward (yaw and roll), or leftward/ downward, rightward/ downward, leftward/ upward, or rightward/upward.

In VE 1 subjects could keep their eyes open if desired, while in VE 2 subjects wore infrared video goggles that covered their eyes and nystagmus was observed with infrared video-oculography to verify that nystagmus was classical [7]. In VE 1 and 2 subjects viewed the six video animations for the test side (see sample video frame in Fig. 3). In VE 2 subjects were also shown the six animations before Dix-Hallpike testing, then proceeded as in VE 1 with the Dix-Hallpike maneuver followed by viewing and selection of videos.

2.4. Compilation and Primary Analysis

For the primary goal, the data were categorized by the number of subjects who could make selections, the number of selections they made, and the combinations of sketches or videos they selected. For analysis of the video data, left and right video selections were combined; e.g., leftward yaw videos for left BPPV patients were counted with rightward yaw for right BPPV subjects, etc.

The main indicator of success of the visual method was given by subjects’ abilities to make selections. A statistical measure was also implemented using data from the last experiment, VE 2. Specifically, we performed a binomial test [38] of the null hypothesis that subjects selected combinations randomly, in the sense of randomly including oppositely-directed pairs (e.g. a subject might select both forward and backward pitch, which is physically impossible). The analysis used data from all subjects who selected at least two animations. The mathematics behind the binomial test is explained in the Appendix. A significance level of p = 0.05 was used in the statistical analysis.

2.5. Comparison with Axis of Eye Movements

Toward the secondary goal, the data from VE 2 were analyzed in comparison with eye movement data. The same method of analysis was used as in Aw et al [2]. Each rotation axis was specified by a unit vector (x, y, z) using the right-hand rule, where x, y and z were the components in the directions of the nose, left ear, and top of the head, respectively. In the videos (Fig. 3), the head was turned 45° in yaw plus 20° pitch back, so the videos for left BPPV, for example, showed rotation vectors for body-referenced pitch = (−0.707, −0.707, 0), body-referenced roll = (−0.664, 0.664, 0.342), and body-referenced yaw = (0.242, −0.242, 0.940). A subject who selected two or more videos was considered to have a rotation axis between those of the selected videos.

The mean perceived rotation axis over all subjects was determined by averaging and then normalizing to again make a unit vector. Per Aw et al. [2], all data from left BPPV were first reflected so that data from both sides could be used in the same analysis. A Student t test for differences between two means of independent observations was used to test for differences between the mean eye movement axis reported in Aw et al. [2] and the mean perceived rotation axis. In addition, the test was also performed by considering that each subject’s indicated axis may be imprecise by as much as 30° because of the discrete choices of videos. The value of 30° amply covers the 22.5° (45°/2) difference between the available reporting combinations because subjects could select two or three videos to indicate a diagonal axis. A significance level of p = 0.05 was used.

For directional data, the Student t test is conservative and is therefore sufficient when detecting statistical significance. A more sophisticated test can be used when necessary, the Fisher-Watson-Williams test [18, 27, 41], which treats angles between axes as multi-directional rather than linear like the Student t test. The Fisher-Watson-Williams test is unfamiliar to most biological scientists, however, and must be coded by hand because no packaged software is available. The Student t test was sufficient for this study.

3. Results

All subjects could select sketches or videos. No subjects had difficulty differentiating pitch rotations from roll or yaw rotations. Some subjects were initially unsure about the difference between roll and yaw rotations but were able to understand with more explanation. Some subjects made their selections immediately. Several subjects viewed the videos twice.

3.1. Picture Experiments 1 and 2

In PE 1 after viewing the five sketches, 15 subjects selected one image, four subjects selected two images, and one subject selected three images. All choices with two or three images included yaw. See Figure 1.

In PE 2 after viewing the five sketches, 12 subjects selected one image, and eight subjects selected two images. One subject selected a diagonal image, but no subjects who selected two images selected the diagonal images. See Figure 2.

3.2. Video Experiments 1 and 2

In VE 1, three subjects selected one video: one each of pitch back, ipsilateral yaw and ipsilateral roll. (Again, the animations showed body-referenced pitch, roll and yaw, so the pitch video represented rotation in the posterior canal plane.) Twelve subjects selected two videos, and ten subjects selected three videos. Thus, 22 of 25 subjects selected videos representing complex motion sensations in more than one plane. See Figure 3.

In VE 2, five subjects selected one video: two backward pitch, one forward pitch, one ipsilateral yaw and one ipsilateral roll. Ten subjects selected two videos, and five subjects selected three videos. As in VE 1, most subjects, 15 of 20, selected two or more videos. See Figure 4.

Figure 4.

Venn diagram showing numbers of subjects selecting each combination of videos in Video Experiment 2. The larger numbers indicate subjects selecting videos with stripes rotating in the typical direction, which this figure shows with arrows, whereas the smaller, “opp.”, numbers indicate subjects whose choices included at least one video with oppositely-directed stripes. Videos are explained in Figure 3.

3.3. Primary Analysis

The success of the visual method was initially demonstrated by the fact that all subjects were able to make selections. Then success was measured more quantitatively by testing statistically against randomness in combining videos, by noting that random selection could include oppositely-directed pairs that are physically impossible in a rotation. The statistical analysis of VE 2 gave the following results by the binomial test (further details in Appendix): 10 subjects selected two animations, and five subjects selected three animations; none of the combinations included an oppositely-directed pair (Fig. 4), resulting in p = (0.8)¹⁰(0.4)⁵ = 0.0011. Therefore, subjects’ selections were statistically significantly consistent with indications of direction of rotation, p < 0.05. Even in the less-refined VE 1, all selected combinations were consistent with directions of rotation; no combinations included oppositely directed pairs as would occur upon more haphazard selection.

Subjects’ selections also made sense qualitatively. Most notably, the selected pitch direction was “backward”, i.e. matched that expected for posterior canal BPPV, in 12 of 14 pitch selections in Video Experiment 1 (Fig. 3), and in 12 of 13 pitch selections in VE 2 (Fig. 4).

3.4. Comparison with Axis of Eye Movements

For comparison with the 3-D axes of eye movements, the mean axis of perceived rotation was determined to be (0.69, −0.14, −0.71) with standard error 12.42°, stated in terms of right BPPV after combining left and right BPPV data. This mean axis had mostly roll (0.69) and yaw (−0.71), with little pitch (−0.14) in head coordinates; instead, the pitch rotation was in Earth coordinates as indicated by rightward roll while the head is turned 45° to the right. This axis was compared with the mean eye movement axis for subjects with unilateral BPPV of the posterior semicircular canal (n = 26) of (0.74, −0.66, 0.17) with standard error 2.15° [2]. This mean eye movement axis had mostly roll (0.74) and pitch (−0.66) in head coordinates. The difference between the axes of perceived rotation and eye movements was 61°, statistically significant at p < 0.0001. Taking into account a wider range of angles (see Methods), the difference between the axes of perceived rotation and eye movements was still statistically significant at p < 0.05.

4. Discussion

4.1. Meaning of results

First, PE 1 and 2 demonstrated that patients with BPPV can use static images to describe their personal perceptions of vertigo. The selection of more than one image by 33% of those subjects suggests that at least some subjects perceived complex motions in more than one plane in space. That idea is supported by the selection of the diagonal picture by one subject in PE 2.

Next, similar to the picture experiments, VE 1 and 2 demonstrated that patients can use video animations to describe their personal perceptions of vertigo. The selection of two or three videos by 88% of subjects in VE 1 and 75% of subjects in VE 2 supports the idea that subjects perceived and could report complex motions in response to the Dix-Hallpike maneuver. The main difference between VE 1 and 2 was an approximate 35% reduction in number of subjects selecting all three backward/ ipsilateral videos in VE 2, suggesting that viewing the videos ahead of time changed the way they reported the sensation. Perhaps they were more precise or already had a context for making the report.

The success of the video reporting method is also supported by the combinations selected. They were shown to be consistent with indications of rotation direction, rather than randomly chosen in a way that includes physically impossible pairs. The data are also consistent with the idea that the perception of rotation is in the same general direction (though not the exact direction, as discussed below) as the quick phases of nystagmus.

A more precise comparison showed that the mean axis of perceived rotation differed statistically from the previously-determined mean for eye movements, when considered as 3-D vectors. This result can be understood intuitively by the fact that, although subjects’ selected videos were consistent with the general direction of eye movements, only a few subjects selected just the top video in Fig. 4 (pitch, in body coordinates), the video that is almost exactly aligned with nystagmus.

Use of these images also had an unanticipated but clinically important effect that we did not measure. The ability to describe the sensation more clearly had a calming influence on some patients. Immediately after their participation in the study, subjects were treated for BPPV by the corresponding author. Many subjects continued to refer to the images during treatment as they described their responses during each phase of the canalith repositioning and liberatory maneuvers [9, 10, 15, 35] as if the images gave them a language with which to explain how they felt.

4.2. Relationship with other research

The comparison with eye movement direction indicates that perception is modified by other factors. For example, the unavoidable presence of the treatment table and the investigator’s hands underneath and supporting the subject’s head might have caused sensory conflict with vestibular input, especially with the body-pitch direction of rotation. Sensory conflict is well known to affect perceived motion, with the somatosensory and vestibular systems contributing in different ways [31]. In this case, the direction with the most conflict—the body-pitch direction—is exactly the direction that is substantially reduced in the perception, as would be predicted by the involvement of the somatosensory system as well as the vestibular system. In this respect, our results in patients mirror those of the healthy subjects tested with caloric stimulation while seated or supine [25]. Among healthy subjects, even though some seated subjects reported perceived pitch, none of the supine subjects did. A common mechanism is likely, involving somatosensory conflict, in the two studies.

The mechanisms of rotation perception are currently poorly understood. Perception undoubtedly includes pathways that are not included in eye movements, although the details of those differences are still under investigation. The time constant of velocity storage changes more for perception than for eye movements upon administration of 4-aminopyridine, a blocker of inward rectifier potassium conductance [36]. The anatomical pathways underlying the effect are not yet clear; areas in the cerebellum are probable [34, 36], as is a vestibulothalamic pathway [23, 37]. An area associated with vestibular perception was identified long ago in the temporoparietal cortex [32]. In addition, convergence of vestibular and somatosensory signals has been found in the cortex [5, 33] and specifically the parieto-insular-vestibular cortex (PIVC) [1, 21]. Since then, several studies have recorded vestibular responses in various cortical areas, including the PIVC, the somatosensory cortex, and other areas, reviewed elsewhere [17, 26]. Multiple pathways have been found by using a combination of structural and functional connectivity mapping in humans [12, 24]. Clearly, however, more research is needed to sort out the differences and mechanisms of perception as compared to eye movements, as well as the contribution of somatosensory input toward the differences.

The current results direct us toward additional research. For example, subjects in the present study often selected more than one video. If these choices were due to the fact that the axes of their illusory rotations, i.e. the axes of vertigo, were between the axes of the available videos, then providing additional videos with a greater selection of rotation axes could help fine-tune subjects’ reports. Further research could also examine whether and/or how precisely subjects perceive a single axis of rotation. Because the present study focused on testing a new method of reporting, a straightforward set of six videos was used here with each subject. Future research, however, could test different combinations of videos and orientations of axes.

4.3. Conclusions

The availability of a video language for dynamic perception has led to both expected and unexpected outcomes. The general agreement found between perception and eye movements confirms the expected outcome, while the mismatch between the more precise axes feeds into a further line of inquiry. Clinically, we have noted some patients’ spontaneous reports of reduced anxiety by using the video language. This video language may be useful in future explorations of the patient’s experience. Furthermore, the reduction of anxiety epitomizes the serendipitous effects of scientific research and may have implications beyond vestibular research.

Appendix

A binomial test is useful when the data set is small and data fall into two categories [38]. Here, a subject’s combination of selected animations either “includes an oppositely-directed pair,” or “does not include an oppositely directed pair.” These are the two categories being analyzed. If subjects tend to choose combinations that indicate a 3-D direction of rotation, then their selections will tend to “not include an oppositely directed pair” because such a pair is physically impossible. (Other combinations are physically possible because the chosen directions can be the components of rotation about an axis oriented diagonally in 3-D.)

To test the null hypothesis that subjects’ selections are random — in the sense that combinations of oppositely-directed pairs occur with a probability indicating randomness — it is necessary to first compute the probability that a single randomly-selected combination includes an oppositely-directed pair. The computations are as follows: When selecting a combination of two animations out of the six available, the probability is 0.2 (= 3/15) because there are 15 (= “6 choose 2”) ways to select two animations, and three of those combinations are oppositely-directed. When selecting a combination of three animations out of the six available, the probability is 0.6 (= 12/20) because there are 20 (= “6 choose 3”) ways to select three animations, and 12 of those combinations include an oppositely-directed pair.

The binomial test uses this information in order to compute the probability that the data set would have arisen if subjects had made selections randomly. If the probability is less than p = 0.05, then we reject the null hypothesis and conclude that subjects did not simply make random selections.

Given the results of the present study, the computation of the probability is relatively simple. Ten subjects selected two animations, five subjects selected three animations, and none of the combinations included an oppositely-directed pair. The probability that this occurred through random selection is the product of the individual probabilities, which are 0.8 (= 1 – 0.2) for two animations, and 0.4 (= 1 – 0.6) for three animations. Therefore, p = (0.8)¹⁰(0.4)⁵ = 0.0011.