We recruited patients over 60 years old from the outpatient clinic for otolaryngology head and neck surgery at the Peking University First Hospital from December 2021 to May 2022. Patients underwent the appropriate vestibular function tests purely for clinical purposes, including: air-conducted cervical vestibular-evoked myogenic potentials (c-VEMP), video head impulse tests (v-HIT), posturography, videonystagmography (VNG) with bithermal caloric tests. All tests were performed by the same clinical technician. Among them, c-VEMP evaluates saccular function, recording from the sternocleidomastoid (SCM) muscles ipsilateral to the stimulated ear in response to a short pure tone (100 dB SPL) delivered monaurally through insert headphones. Take a 10 dB step until no recognizable P1 and N1 waves (first positive and negative wave with latency ranging from 13 to 23 ms) can be seen. The lower frequencies function of the lateral semi-circular canals was evaluated by v-HIT and bithermal caloric tests. During the v-HIT test, subjects were 1.2 m away from the visual target. Vestibulo-ocular reflex (VOR) gain is defined as the ratio of the angular velocity of eye movement to the angular velocity of head movement. Caloric irrigation was conducted by using cold air at 24°C and hot air at 50°C. Each irrigation was performed after the evoked nystagmus had completely disappeared. Patients who met the following criteria were included in the group of older vertigo patients (OVP).

Inclusion criteria for clinical patients were:

(1) Had a clinical history of recurrent vertigo.

(2) Failed at least one of the following vestibular function tests:

(3) Subjects might suffer from a post-lingual hearing loss.

To better illustrate the effect of age on visuospatial ability, three types of control participants were recruited, including a young group (YC; 18–44 years old), a middle-aged group (MAC; 45–59 years old), and an older group (OC; 60 years old and above). Some of them were recruited from society, and some were from the examiner's friends and hospital staff. The inclusion criteria for the control group were: (1) 18 years of age and above; (2) had no history of benign paroxysmal positional vertigo (BPPV), meniere disease (MD), vestibular neuritis (VN), vestibular migraine (VM), and other diseases that may cause vertigo.

In addition, audiometric data were also available from subjects using a clinical audiometer (AD229e, Interacoustics, Denmark), TDH39 headphones, and testing in a standard soundproof booth (< 30 dB A). Hearing status was indexed by the averaged pure tone hearing threshold (PTA) of the four frequencies (0.5/1/2/4 kHz) of the better ear.

Basic information for all subjects, including gender, age, and education, was collected before the test. For both patients and controls, individuals were excluded if they: (1) had hearing loss that affects daily communication; (2) had visual impairment; (3) had middle ear disease or long-term noise exposure; (4) had a history of psychiatric and/or neurological disorders such as anxiety and depression; (5) had years of education < 6 years; (6) had dementia disease, such as Alzheimer's Disease (AD).

We used the Lenovo TB-J606F tablet with a resolution of 2,000 × 1,200 and a screen size of 11 inches to run the VCAS, which includes four modules: the “Weeding Test,” “Maze Test,” “Three-dimensional Driving (3D Driving),” and “Card Rotation Test.” Subjects were seated next to the experimenter with the tablet in front of them and completed tasks by touching the screen. Before any formal test, the subject is first directed through the “training mode” designed for each test to make sure that they have familiar with the processes. To avoid learning effects, the items in the training mode are different from those in the formal test. It takes approximately 40 min to complete all the tests. For patients, visuospatial assessment was performed before vestibular function tests. All test results will be stored in the Tencent Cloud storage bucket and displayed on the “Result Query” page allowing further analysis.

We designed the “weeding test” inspired by the traditional Corsi Block Tapping paradigm and the work of Claessen et al. (31) and Lacroix et al. (32), which evaluates visuospatial attention and working memory processes. There are two sub-tests in the task: a forward and a backward condition. During the test, 9 squares symbolizing the “grass” were shown on the screen with a flashing time of 500 ms and an inter-block interval of 1 s. The relative block positions were the same as the traditional paradigm. Subjects have to memorize the sequence and click on the corresponding “grass” (i.e., weeding). In the forward condition, subjects need to reproduce the block sequence in the same serial order as indicated by the system, while in the backward condition, they repeat it in reverse order. Two trials per sequence length, ranging from two to nine blocks (for backwards, ranging from two to eight blocks). As long as the subject reproduce the block sequence completely correct in either of the two trials, one can enter the next sequence, and the game is over if the subject fails twice. The system automatically registered performance in terms of the span (forward/backward) which is used to evaluate the spatial working memory and the velocity (forward/backward), which is used to reflect the subject's click speed during the working memory process. The velocity was defined as the ratio of the total number of blocks clicked by the subject to the time spent in the test. The diagram is shown in Figure 1A.

Inspired by Lacroix et al. (32), we designed the “maze test” to assess spatial navigation and executive function. The map was randomly generated through math algorithm, which can be divided into numerous small squares (map “8 × 8,” i.e., this map can be divided into 64 small squares). The complexity of the map depends on the number of squares. Users can create different mazes for specific purposes by entering specific numbers in the input box at the top right of the screen and clicking the button “Create a maze.” According to the results of pre-experiment in a normal population, we included three maps of different difficulty levels (8 × 8, 10 × 10, and 12 × 12) here. At the subject level, the degree of difficulty is controlled by the number of corners. The calf in the lower left corner of the screen is the “starting point” and the straw in the upper right corner is the “destination.” Subjects have to move the calf to the “destination” by clicking the control panel in the lower right corner of the screen. There is only one correct route. All the subjects were instructed to get out of the maze as quickly as possible. The system automatically registered performance in terms of the time subjects took and the steps they moved for each maze. The diagram is shown in Figure 1B.

Spatial memory and spatial navigation are often interdependent in real life. Inspired by Coutrot et al. (29), we designed “3D driving” to assess spatial memory and spatial navigation in a comprehensive manner. When the test begins, a two-dimensional map with all the intersections will be displayed in the center of the screen for 5 s. The subjects need to keep the map in mind and “drive” the car to the destination in a virtual three-dimensional scene according to the memorized route. Taking into account the limited operation ability of older people, we designed the autopilot mode to make the car move automatically and stop at each intersection. Only if the subjects click the buttons for direction selection (“turn left,” “forward,” and “turn right”) that appear on the screen correctly, the car will continue to drive. If the wrong selection was made, one will be prompted to re-select. To avoid learning effects, subjects have to complete three times (maps with different routes only) and the final results are averaged. The program automatically registered performance in terms of the thinking time and the number of errors at intersection. The diagram is shown in Figures 1C,D.

The “card rotation test” is inspired by traditional mental rotation paradigm and the work of Lacroix et al. (32) and Henn et al. (33). There are 8 questions in total. Each question takes a 2D or 3D figure as an example, and at the bottom of the screen are four figures with the same size and color but different rotation angles from the example. We used letters and numbers as 2D-object and blocks with numbers as 3D-object. Subjects were required to mentally switch to determine the option that was exactly the same or completely different from the example and click the corresponding button. There is only one correct answer. The same difficulty level included 2 questions. A correct answer is scored as 1 point, and an incorrect answer is scored as 0. The program automatically registered the performance in terms of the final score and the time spent on the subject. The diagram is shown in Figures 1E,F.

All statistical analyses were performed with SPSS 25 (IBM; Armonk, NY, United States). The normality of distributions was evaluated using Shapiro-Wilk tests. Demographic data were analyzed with parametric analysis of variance (one-way ANOVA) for continuous data and chi-squared test for categorical data. We performed a multivariate ANOVA (for the weeding test and maze test) or multifactorial ANOVA (for the 3D driving test and card rotation test) followed by a post-hoc test (LSD) to compare the performance of different groups on each test of the VCAS. Each index was used as the dependent variable. Group, education, and gender were used as fixed factors, and only the factors with main effects were tested for interaction effects. For data do not meet the normal distribution, a ln logarithmic transformation is performed before ANOVA. To compare differences between subjects in the weeding test, two-way analyses of variance (ANOVAs) for repeated measures were performed on the span with 4^*group (OVP, OC, MAC, YC) as a between-subject factor and 2^*recall order (forward, backward) as a within-subject factor. The effect sizes are reported in terms of partial η². An alpha level of 0.05 was used.

A total of 184 subjects participated in the study. Of these, 67 were clinical patients, and 117 were controls. According to the inclusion and exclusion criteria, 24 patients were excluded, including 15 with incomplete basic information, and 9 had normal vestibular test results (1 with a right ear perforation). Two controls with cataract were also excluded. Finally, 158 subjects (63.92% females) were included, including 43 patients (OVP; mean age: 66.14 years, SD: 4.5), 32 older controls (OC; mean age: 66.06 years, SD: 5.95), 38 middle-aged controls (MAC; mean age: 50.87 years, SD: 4.10), and 45 young controls (YC; mean age: 34.02 years, SD: 6.52). According to education, subjects were divided into primary education; secondary education, which includes middle school, high school, and technical secondary school; and higher education, which includes junior college, university, and postgraduate.

Age differences between the groups were statistically significant (p < 0.001). Post-hoc analyses indicated comparable ages between the OVP and the OC group. Gender differences were not statistically significant (p = 0.190). Education differences were statistically significant (p < 0.001). Post-hoc analyses showed no statistically significant difference between the OVP group and the OC group. Audiometric data was available in 116 of 158 subjects. The difference between the groups was statistically significant (p < 0.001). Post-hoc analyses showed that it was comparable between the OVP and the OC group as well as between the MAC and the YC group. Since the hearing thresholds of the OVP and OC groups were almost close to the normal range ( ≤ 25 dB HL), the hearing performance of the subjects was not further examined in this study. Demographic and hearing performance are presented in Table 1. Results of the vestibular function tests and hearing results of patients are shown in Supplementary Table S1.

For the weeding test, results displayed in Figure 2 and Table 2 show that only the factor group had a significant main effect on the span (F_FW = 7.062, p < 0.001, η²_p = 0.135; F_BW = 3.406, p = 0.020, η²_p = 0.070). Post-hoc analyses showed that the forward span was significantly shorter in the OVP group than in other groups (p < 0.001). For the backward span, there was no statistically significant difference between the OVP and OC groups (p = 0.057), but the OVP group performed significantly worse than the MAC group (p = 0.046) and the YC group (p < 0.001). Regarding the velocity, there was a main effect of group on backward condition (F_FW = 1.343, p = 0.263, η²_p = 0.029; F_BW = 2.709, p = 0.048, η²_p = 0.057), and a significant main effect of education on both forward and backward conditions (F_FW = 5.633, p = 0.004, η²_p =0.078; F_BW = 5.063, p = 0.008, η²_p =0.070). Post-hoc analyses showed the differences between the OVP and OC groups were nonsignificant in both conditions (FW, p = 0.654; BW, p = 0.527). Between the OVP and MAC groups, only the difference in the backward condition was significant (p = 0.043). Additionally, the OVP group performed significantly worse than the YC group in both conditions (p < 0.001). People with higher education levels had a faster click rate (FW, p = 0.004; BW, p = 0.008). The interaction between group and education was nonsignificant (F_FW = 0.817, p = 0.558, η²_p =0.055; F_BW = 1.008, p = 0.422, η²_p =0.057).

Results of the two-factor repeated measures ANOVA revealed significant within-subjects effects (Figure 3). The forward span was significantly longer than the backward span (F = 21.159, p < 0.001, η²_p = 0.121). There was no interaction effect between group and recall order (F = 1.773, p = 0.155, η²_p = 0.034). That is, the difference between different recall orders did not change with the group. A paired t-test applied to the OVP group individually revealed no significant difference between the different recall orders (t = 0.461, p = 0.648).

Data in the maze test was transformed by the ln function before ANOVA. Results are shown in Figure 4 and Table 3. In all mazes, group had a significant effect on time [F_{(8 × 8)} = 5.995, p = 0.001, η²_p = 0.130; F_{(10 × 10)} = 7.666, p < 0.001, η²_p = 0.161; F_{(12 × 12)} = 9.690, p < 0.001, η²_p = 0.195]. Post-hoc analyses showed that the OVP group took significantly longer time than the OC group in different mazes except the maze 10 × 10 (8 × 8, p = 0.009; 10 × 10, p = 0.953; 12 × 12, p = 0.032). Gender had a main effect on the time for maze 10 × 10 and maze 12 × 12 [F_{(10 × 10)}= 5.476, p = 0.021, η²_p = 0.044; F_{(12 × 12)} = 12.264, p = 0.001, η²_p = 0.093] expect the maze 8 × 8 [F_{(8 × 8)} = 2.662, p = 0.105, η²_p = 0.022]. There was no interaction effect between group and gender in all mazes [F_{(8 × 8)} = 0.004, p = 1.000, η²_p < 0.001; F_{(10 × 10)} = 0.104, p = 0.957, η²_p = 0.003; F_{(12 × 12)} = 0.212, p = 0.888, η²_p = 0.005]. Regarding the steps of the maze, no factors included in the analysis had main effect.

Results of the 3D driving test are shown in Figure 5 and Table 4. There was a main effect of group and gender on thinking time (F_group = 5.177, p = 0.003, η²_p = 0.195; F_gender= 5.472, p = 0.022, η²_p = 0.079). And there was no interaction effect between these two factors (F = 0.581, p = 0.630, η²_p = 0.027). Post-hoc analyses showed that the OVP group took longer time to think than the OC group, while the difference between the groups was nonsignificant (p = 0.774). Regarding the number of errors, there was only a main effect of group (F = 4.514, p = 0.006, η²_p = 0.151). Post-hoc analyses showed that the OVP group made a few more errors than the OC group, but the difference was nonsignificant (p = 0.399).

Results of the card rotation test are shown in Figure 6 and Table 5. There was no main effect of group on both score (F = 0.434, p = 0.729, η²_p = 0.021) and time (F = 0.157, p = 0.925, η²_p = 0.008). However, the OVP group indeed scored lower and took more time than the OC group, although the difference between the groups was nonsignificant (p = 0.400), while the difference in time between the groups was nearly statistically significant (p = 0.056). The results showed that only education had a main effect on the score (F = 3.190, p = 0.048, η²_p = 0.093). Post-hoc analyses showed that subjects with higher education scored a little higher than the other two groups.

For all the test metrics, we observed a stepwise increase in the performance between young controls, middle-aged controls, and older controls, although there were no significant differences between the OC group and the MAC group (Supplementary Figure S1). The YC group performed significantly better than the other groups separately in the span of the weeding test (FW: p = 0.001, p = 0.004; BW: p = 0.015, p = 0.011), the velocity of the weeding test (FW: p = 0.001, p < 0.001; BW: p < 0.001, p = 0.017), the time of the maze test (8 × 8: p < 0.001, p < 0.001; 10 × 10: p < 0.001, p < 0.001; 12 × 12: p < 0.001, p < 0.001) and the thinking time of the 3D driving (p < 0.001; p = 0.001). For the number of errors in the 3D driving, only the OC group had significantly more errors than the YC group (p = 0.014). In the card rotation test, although there was no main effect of group on both score and time (F_score = 0.434, p = 0.729, η²_p = 0.021; F_time = 0.157, p = 0.925, η²_p = 0.008), post-hoc analyses revealed that the YC group scored significantly higher than the OC group (p = 0.005) and the MAC group (p = 0.006), but the differences in time between all the control groups were nonsignificant.

Spatial memory stores and manages information about the environment (e.g., location and relative position). We designed a computerized version of the Corsi block tapping task (CBT) (31) to evaluate visuospatial working memory (VSWM). Results found that the OVP group had a significantly shorter span than the age-matched controls in the forward condition. This is in line with previous work that short-term spatial memory was impaired in patients with chronic vestibular dysfunction during computerized CBT, with 4.11 ± 1.07 and 5.29 ± 0.77 for the forward span of older patients and controls, respectively (34). Unfortunately, they didn't talk about the backward condition. Our results showed that the OVP group indeed performed worse than the OC group in the backward condition, but the difference was not statistically significant, which may be due to the backward condition being more difficult and affecting the performance of controls. Regarding the differences between the forward and backward conditions, some studies argue that the backward condition, in which participants need to recall in the opposite order, increases the burden on cognition and is therefore more strenuous (27). However, similar results have been found between these two conditions (31), and one study even found that subjects performed better on the backward condition (35). In the present study, the forward span was significantly larger than the backward condition for controls, supporting the very first perspective. However, difference of the performance in the OVP group was nonsignificant, this may be due to the patients' poorer performance in the forward condition. There is still no consensus on whether these two conditions indeed tap into different cognitive processes. Future work is needed to further explore this issue.

For spatial navigation ability, results found that the OVP group took significantly longer time than the OC group. We did not observe a significant difference in steps between the two groups. It should be noted that during the test, we observed that more subjects in the OVP group took the wrong route and then retraced it to the correct one, which may lead to a significant increase in the time and steps. This can also be reflected on the larger interquartile range (IQR) of the OVP group. Our results confirm and extend prior reports regarding poorer performance in spatial orientation and spatial navigation in patients with vestibular dysfunction (9).

During the 3D driving test, although there were no statistically significant differences, the OVP group indeed made more errors than the OC group, suggesting that patients may have had some difficulties, such as poorer memory of routes or being more susceptible to getting confused. However, this did not seem to affect their ability to make the correct choice they thought. It is also to be noted that there were only three intersections in the map, which may have weakened the differences between groups.

As to mental rotation ability, we also did not find statistically significant differences between groups, although the OVP group still had a lower score and took longer time. This may be due to the highly diverse in terms of the type of vestibular diagnosis among patients, which was not sufficient to reflect the differences with controls in this task. Currently,the research on mental rotation ability is still controversial. Some studies have found that, BVP patients and patients with VN or BPPV may exhibit significant impairment in mental rotation task (10, 16). However, Nair et al. (36) found no significant differences in overall performance between BPPV patients and controls in a mental rotation task of 3D objects. More studies with large samples will be helpful for this issue in the future.

With age, spatial orientation and navigational abilities may deteriorate (26). Collectively, in our study, the YC group had a better performance than other groups which is consistent with previous studies. However, the OC group and the MAC group had comparable performances during all the tests. We infer that visuospatial ability decreases with age and may gradually stabilize thereafter. It is intriguing that in the study by Coutrot et al. (29), there is an “inflection point” where performance on the navigation task improved for those over 75 years of age, and the authors explain that this could be a selective bias that these people may have extraordinary cognitive abilities.

The solid influence of aging on visuospatial abilities can be explained in terms of fluid intelligence (FUI), which refers to the ability to transfer and reason independently of prior experience and knowledge, and is more susceptible to aging than crystal intelligence, which relies on inherent experience accumulation (37). In addition, vestibular loss associated with normal aging may also mediate age-induced decline in cognition (38, 39). Therefore, it is not yet able to fully exclude the effects of age-related vestibular loss in this study. In the future, multicenter studies focusing on the effect of aging on visuospatial ability with larger sample size and finer age group delineation can help further clarify this issue.

Previous studies have used a variety of instruments to assess visuospatial ability such as comprehensive scales or single-dimensional neuropsychological tests. Most of the tools are paper and pencil tests, which have some shortcomings: (1) It is difficult to control the experimental conditions accurately and is inconvenient for the management of test data. (2) The test conditions are mostly static, with little sense of dynamics. (3) Concentrate primarily on one dimension of visuospatial abilities. Furthermore, although some paradigms already have a computerized version, such as the computerized CBT (27, 31, 40), most of them are limited to the PC, which is inconvenient to carry for bedside evaluation. Therefore, we hope to develop an effective and convenient tablet-based test battery to provide a more dynamic and three-dimensional test condition and mainly focus on evaluating visuospatial ability, including spatial memory, spatial navigation, and mental rotation.

Our test system has specific advantages like other computerized task, such as precise control of the presentation interval and automatic recording of test metrics. However, we should be aware that there are still some potential issues. Firstly, it should be noted that there are possible differences between computerized versions and traditional test paradigms. For example, during the CBT, Popp et al. found a statistically significant difference between BVD patients and controls in the backward condition using a traditional tapping paradigm (41), which is different from the results we observed. One study has proposed that the hand movements of the experimenter in the traditional paradigm induce a motor-priming effect (31). The motor stimulus pre-presented will have either a positive or negative effect on the observer's motor processing, which indicates that the mental processes involved in the two modalities may differ and may limit the generalization of the results. Collectively, at least in the forward condition, there is an effect of vestibular dysfunction on spatial memory, and the use of the backward condition as a representative of spatial memory should clearly be reconsidered, especially in such computerized version. A larger number of subgroups using different modalities of CBT will be helpful to systematically investigate this issue.

Secondly, it has also been proposed that subjects do not rely on vestibular input from actual movement when completing such computerized task and only need to remain seated during the test, and attentional resources can be directed exclusively to the spatial memory task (42). Therefore, a purely static test setup may not be sufficient to capture the degree of impairment in real-life in patients with vestibular dysfunction. However, Kalová et al. (30) found that the results of a navigation test using a computerized version were similar to navigation performed in the real world, supporting the use of a computerized test modality to assess cognitive function.

In addition, the impact of subjects' familiarity and operation ability with electronic devices should be carefully considered when applying such computerized versions of the test, especially for older adults. To sum, although a computerized test system has been proven to have certain practical advantages, it is more sensible to be careful when quantitatively comparing visuospatial ability in virtual and real-world environments. Computerized versions of neuropsychological tasks should be carefully evaluated through a large-sample research design with normative data established for different age groups as well. And we believe that combine tablet-based test system and VR technology may improve our assessment tools more portable and effective in the future.

Of note, there are several limitations of the study. First, the cross-sectional design limits the causality of conclusions that can be drawn from it. Future work is needed to further explore such scenarios by developing longitudinal cohort studies; Secondly, the sample size of the subgroups in this study was small, which makes it difficult to further investigate the disease-specific effect within the OVP group and demonstrate a more refined trend in visuospatial abilities of controls; Finally, although the computerized version has certain advantages, it still lacks the input of dynamic vestibular information, which may not be sufficient to reflect the cognitive symptoms of patients.