Vestibular Aging Process from 3D Physiological Imaging of the Membranous Labyrinth

There is no three-dimensional (3D) technique to study the microanatomical structures of the in vivo 3D vestibular membranous labyrinth. Recent two MRI methods using a contrast agent can only depict the low-resolution imaging of endolymphatic hydrops. Therefore, we provide the new precise volume rendering algorithms to create the in vivo 3D vestibular membranous labyrinth images from high-resolution temporal bone low-dose CT data. We also ascertain whether the created 3D microstructure images are reliable in anatomical findings. Secondary, we will analyze the age-related changes of the vestibular membranous labyrinth. These created 3D membranous vestibular images were almost consistent with the appearance, dimensions, areas, and angles from those acquired in previous histological works. The age-related image changes showed the enlarged saccule in females, the enlarged utricle in males, and the dilated tendency of the lateral semicircular duct. These results may correlate to the findings of the previous physiological works on cervical and ocular vestibular evoked myogenic potentials, and gait studies. The age-related balance disorders may be associated with the enlargement of each membranous organ in the vestibule. This new imaging technique now enables visualizing microanatomical changes in the in vivo membranous vestibulum, and these created 3D images may suggest physiological information.

The Angle between utricle and saccule. It is no statistically significant association as NS (t = 0.385, f = 4.071, p = 0.720).

Analysis of the age-related changes in the vestibular membranous labyrinth.
There was no correlation between the membranous CT value and the age group in gender differences as follows: both genders; r (r 2 ): 0.172 (0.030), males; −0.194 (0.038), and females: 0.125 (0.016) The statistical results of the age-related changes in each organ of the vestibular membranous labyrinth were shown in Table 3. The summarized results were as follows.
Utricular area. There was a statistical relationship in males, but not in females.

Saccular area.
There was a statistical relationship between both genders. Females suggested a relationship as p = 0.056.
The angle between the utricle and the saccule. There was a statistical correlation in both genders.
The width of the lateral semicircular duct. A significant statistical correlation was in males, but a very weak correlation in females.

Figure 1.
Anterior cut view of the utricle and the saccule in a 60-year-old female. This image is created from an upward curve with a merged color algorithm under 40% opacity value. The anterolateral aspect of the utricle is seen in the elliptical recess. The saccule is located in the spherical recess in the anteromedial part of the vestibule at the level of the oval window. The anterior ampulla connected with the utricle looks like a bud shape. The running nerves can be seen in this image. The dimensions of the utricle are measured as in this figure. Abbreviations: ut = utricle, sa = saccule, sap = superior ampulla, ssd = superior semicircular duct, rd = ductus reuniens, ow = oval window, er = elliptical recess, sr = spherical recess, utn = utricular nerve, vg = vestibular ganglion, san = saccular nerve.

Discussion
The created 3D vestibular membranous labyrinth image is almost consistent with the previous histological ones because of its shape and dimensions. There is a statistically significant difference in the length of the utricle between our measurements and the previous histological ones. However, we selected seven higher measurement results (a mean ± SD: 2.73 ± 0.05 mm; n = 7), and the statistical difference test was performed again. The Welch test showed NS (t = −2.325, f = 6.382, p = 0.057). Therefore, these generated 3D microstructure images can be considered to approximate the in vivo membranous labyrinth. The precise volume rendering technique is suitable to delineate the complex microanatomy of the membranous labyrinth. It depends on the optimal algorithms. Therefore, the different results can be obtained from them and the created microanatomical images appear qualitatively different. These algorithms provide a translucent view, thus giving a good impression of the spatial relationships of all microanatomical structures. For the optimal imaging result, it is necessary to use the different visualization techniques for each anatomical tissue in the membranous labyrinth.
In a threshold method, volume rendering causes tissues to fade from visible to invisible when thresholding completely hides or shows tissues of a certain density. However, partial volume effects can generate a phantom voxel value within the threshold range selected at the interface between dark and bright regions. In this study, the partial volume effect does not need to be considered, as the osseous vestibular values (a mean ± SD: 284.65 ± 47.80HU) were over the threshold value (under 160HU).
However, this technique has some problems with an algorithm, a threshold and, an opacity. There is a positive statistical correlation between the membranous CT density and the opacity values. So, the optimal opacity curve algorithm will fit the upward curve. However, to create all 3D membranous structures, we need to use the optimal opacity curve algorithm separately for each microanatomical structure. That is, creating different 3D microanatomical images requires the use of different combined algorithms separately. And more, to create the precise 3D microanatomical image, the optimal opacity should be adjusted to the optimal threshold for the target organ. These variable values also will lead to more precise microanatomical information. At this point, further study is needed.
The second problem is image resolution. In these minute microanatomical areas, image resolution is the most concerning issue. The resolution in this study is 0.18 × 0.18 × 0.5 mm 3 . This value is relatively rough for www.nature.com/scientificreports www.nature.com/scientificreports/ describing the microstructures of the membranous labyrinth. Therefore, a smaller FOV and thinner slice thickness should be used. However, this problem is that it depends on the capabilities of the CT machine.
The third problem is the depiction of the otoliths from this technique. It is impossible to create the otoliths using these created algorithms. Because, the otoliths are made of calcium carbonate (CaCO3), and the CT value of CaCO 3 exceeds 140 HU (a mean ± SD: 189 ± 38.4) 29 . This value is higher than the membranous vestibular CT value (a mean ± SD: 110.446 ± 21.176 HU). So, we cannot create the otoliths from the algorithms under the defined threshold in this study. When investigating the otoliths, we need to create other algorithms with the optimal threshold value. The part of this study was published before as an abstract form 30 . However, this technique is relatively simple and safe, and can be used for daily practice to know the membranous labyrinth. Because this technique using a conventional CT machine is low radiation dose exposure. The DLP was 40 to 55 mGy/cm. That is, the radiation dose was 0.06 to 0.08 mSv 31 . This radiation dose is about the same as a skull X-P 32 .
On the other hand, endolymphatic hydrops imaging from two MRI methods has greatly impacted the field of otology and has prompted an ongoing shift in the diagnostic paradigm for Meniere's disease (MD) [3][4][5] . These MRI methods using GBCAs with a head coil can obtain only a bidimensional image of endolymphatic hydrops. Therefore, these methods have three disadvantages.
The first concern is the potential for toxicity of GBCAs. The mechanisms of GBCAs toxicity is gradually revealing from recent studies [6][7][8] . Rogosnitzky & Branch 6 review gadolinium accumulation in various tissues of patients without renal impairment, including bone, brain, and kidneys. One of the most important age-related pharmacokinetic changes is decreased renal excretion of the drug. Older people generally have less muscle volume and less physical activity than younger adults, and therefore less creatinine production. So, their serum creatinine levels usually remain within the normal range despite a reduced glomerular filtration rate (GFR) in the elderly. Therefore, the contrast agent administration will be considered carefully, even if normal serum creatinine levels are maintained. Rah et al. 7 state that intratympanic contrast injection can cause local toxicity in animal models. However, it may not be likely that a GBCA will lead to hair cell toxicity, as the estimated concentration in the inner ear after clinically tried intratympanic injection is far more diluted. Type I and type II hair cells decrease with age and are fitted by a linear regression model 33 . Hence, this method in the elderly may decrease hear cells. The newly added alert from the FDA in 2018 states that gadolinium can stay in the body for months to years after receiving these drugs during an MRI scan 8 . It is important to note that metabolism declines with age. If GBCAs administration is the best option for a given patient, a pretreatment approach that protects against toxicological endpoints may be a viable option. www.nature.com/scientificreports www.nature.com/scientificreports/ The second concern is the low-resolution images of the endolymph or perilymph within the minute structures in the inner ear 3,4 . The image voxel volume for these methods is approximately 0.5 mm 3 . It is not suitable to visualize micro-changes in the vestibular membranous labyrinth. We think that this factor makes the inability to distinguish the two adjacent structures between the utricle and the saccule because of the same signal strength and no visible boundaries between them. Therefore, it is probable that the vestibular end-organs cannot be measured accurately than our method.
The third concern is whether these methods can be used as a new test for the symptom-based diagnostic criteria for MD 34 . Lopez-Escamez & Attyé 5 review magnetic resonance imaging for the diagnosis of MD. Their significant concept from the imaging of endolymphatic hydrops is the realization that clinical MD likely only represents a partial match of patients with endolymphatic hydrops and clinical symptoms. Because the patients for the typical clinical definition of MD often do not fit the endolymphatic hydrops imaging. They state that endolymphatic hydrops can be reliably measured in the saccule using the saccule to the utricle area ratio (SURI). This implies that the saccule plays a major role in MD. The saccule may not expand in some of MD cases. The reason for this discrepancy is that the ability to measure a microanatomical structure does not depend on the patient's symptoms but technical factors. Therefore, these MR methods cannot use as a new diagnostic tool for MD. To solve this problem, it is necessary to create high-resolution MR imaging techniques. However, their review sheds light on the complexities of MRI in the diagnosis of MD but it is clear more research is needed, including the use of our method.
Our measurement results imply the age-related changes in the vestibular membranous labyrinth. However, there are no age-related changes in the membranous vestibular CT value in both genders. In other words, the CT value of the vestibular substances is almost constant regardless of age. On the other hand, the dimensions of the utricle, the saccule, and the lateral semicircular duct show the age-related changes. And these changes suggest that they may correlate with some of the changes in physiological aging. The angle between the utricle and the saccule increases with age. This means that the vestibular membranous labyrinth will deform in each organ with age, and then the angle between them will increase. In this study, the saccule expands with age in females, and the utricle expands with age in males. These results imply that these minute structural changes may be correlated with the physiological findings from the several previous articles of ocular and cervical vestibular evoked myogenic potentials (VEMPs), and of gait studies [12][13][14][15][16][17][18] . The origin of cVEMP is considered to be the saccular function, and the origin of oVEMP is considered to be the utricular function. The elderly ocular and cervical VEMPs amplitude are small. cVEMP response latencies are also longer and depend on a greater extent on stimulus volume to www.nature.com/scientificreports www.nature.com/scientificreports/ generate an effective response in the elderly. oVEMP response rate has been shown to decrease with normal aging. The elderly show less ocular counter roll during slow roll tilt and also in response to electrical vestibular stimulation consistent with decreased utricular responsiveness. Although all studies find that the latency increases  www.nature.com/scientificreports www.nature.com/scientificreports/ with age, females may be less likely to show this latency change than males 12,13 . These physiological findings may suggest a correlation with our measurement results that the utricle expands with age in males, not in females. Regarding walking, the decrease in walking speed with age is slower in males than in females 14 . Makiura et al. 15 reported that stride-to-stride time variability (STV) was significantly increased in females compared to males, and females exhibited significantly higher root-mean-square (RMS) of vertical and anterior-posterior axes and lower auto-correlation coefficient (AC) of vertical axes than males. These results suggest that females have lower gait stability than males. The reactive control mechanism, distinct from the proactive control balance mechanism, mainly depends on somatosensory and vestibular systems to determine the extent and the stimulation type and to trigger a properly scaled postural response 16 . Females have lower AC of vertical axes than males. This phenomenon is interpreted as being associated with a relatively saccular insufficiency in females. Although age-related vestibular losses are noted for both genders, oVEMPs are not associated with gait speed in both genders 17 . These results suggest that gender differences in walking speed with age may be caused by the saccular function. These suggest a correlation with our results that the saccular dimension with age increases in females but not in males. For the lateral semicircular duct, investigations on the effects of normal aging on caloric response find a significant increase in response for middle-aged groups followed by a slow decline with increasing age 18 . In this study, the width of the lateral semicircular duct in the vicinity of the ampulla is significantly related to male aging and very weakly related to female aging. Thus, it will imply that the endolymphatic flow velocity in the semicircular duct slows with age due to its dilation.
These created in vivo 3D vestibular membranous labyrinth images have the potential to reveal the interrelationship of each organ in the in vivo vestibular microanatomy. The in vivo 3D microanatomical image changes may resemble the results of previous physiological works. Therefore, these created images may be useful for the study of physiological phenomena in the vestibular membranous labyrinth.
In conclusion, the 3D microanatomical imaging technique can be used for the precise microanatomical study of in vivo 3D vestibular membranous labyrinth. The micromorphological changes demonstrated by this method may represent the age-and gender-related changes in balance. The created 3D microanatomical image suggests physiological phenomena. We believe that this 3D technique will improve the understanding when dealing with various etiopathology and treatments of the vestibular end-organs.

Participants. This study was conducted under the approval of Tanioka Clinic review board with the 1964
Helsinki Declaration and its later amendments or comparable ethical standards. All participants gave written informed consent.
Imaging data from 20 normal ears of 10 participants (10 ears of 5 females, mean age: 53.6 years; 10 ears of 5 males, mean age: 45.6 years) who visited for their regular check-ups and wished to inspect the head and temporal bone regions were obtained. These participants had no known affliction of the temporal bone and no clinical history of hearing and balance problems, and normal findings on both CT exams and physical exam by their check-ups. For histological subjects, the histological results of previous works of literature were used 19-28 . CT protocol. All examinations were performed with a spiral CT scanner (ProSpeed AI; General Electric Systems, Milwaukee, Wis, USA). The section thickness was 0.5 mm. A voltage of 120 kV and a current of 60-120 mA were used. The axial images of the temporal bone were reconstructed with a high-resolution bone algorithm in steps of 0.5 mm and a FOV of 96 × 96 mm 2 using a 512 × 512 matrix.  Table 3. Simple linear regression and correlation between each organ measured value and age.
www.nature.com/scientificreports www.nature.com/scientificreports/ Volume rendering and opacity value. Volume rendering can be used to show both objects at the same time. We seek to display a membranous tissue among the surrounding osseous vestibule. By assigning a low opacity to the osseous vestibule and a high opacity to the membranous tissue, the membranous tissue becomes clearly visible among its semi-transparent surroundings. Where thresholding either entirely shows or hides tissue of a given density, volume rendering shows the tissue to fade from visible to invisible. Assigning the same density to both makes them indistinguishable. When the density of the outer object gradually decreases, the inner object becomes visible. That is, the density ratio between them is the highest opacity.
In this study, the opacity value was calculated by the formula.
= × Opacity value(%) membranous density(HU)/osseous density(HU) 100 Opacity curve algorithm. Tissue-specific images with a model defined in terms of opacity were created by various opacity curve algorithms. We created optimal combination algorithms suitable for imaging each membranous organ as follows [9][10][11] .
A. An upward opacity curve using a merged color algorithm An upward curve algorithm shows bright voxels. It may be used to display bright structures, such as bone or vessels in CT data sets. Therefore, this is suitable for the soft tissues on the osseous lamina. B. An upward opacity curve combining a merged color and a step algorithm A step curve algorithm displays a surface shading image. This is a step curve with only one control point. It allows us to obtain a surface shading rendering on the volume rendering views. This algorithm fits imaging semicircular ducts. C. A trapezoid opacity curve using a non-merged colored algorithm A trapezoid curve algorithm shows voxels within a threshold range. This curve type is used to display structures with voxel values within a range. This algorithm is common for creating the membranous labyrinth. D. A downward opacity curve using a merged colored algorithm A downward curve algorithm shows dark voxels. This curve type can be used to display dark structures, such as the organs of Corti within the cochlear duct, the saccular, utricular, and ampullary organs.
Setting volume rendering parameters. Using a transparent volume-based technology, we defined parameters to carry on the algorithms from a threshold method. If the brightness value was set constant under the defined threshold value, the created image depended on the opacity value. In this study, the brightness value was defined as the constant value of 100%. Therefore, to create a 3D image, we must determine the optimal threshold and opacity values from vestibular CT values.

Measurements of vestibular CT Value.
To define the optimal threshold value, the radiodensity of both the whole vestibule along with the bone wedge (osseous vestibule) and the soft tissue component (membranous vestibule) within the osseous vestibule were measured in Hounsfield units (HU) using 0.5-mm-thick axial high-resolution temporal CT image slices at the level through the oval window, the cochlea, and the vestibule. The imaging conditions were 2000 HU in window width, and 200 HU in window level. The membranous vestibule was composed of the soft tissues, the nerve fibers with the myelin containing the lipid component, the fluids, and the fat. Therefore, the lowest CT value was fat 35 of the lipid component. So, the lowest threshold and opacity values were −100 HU and 0%, respectively. Table 4 showed the measurement results and opacity of two parts of the vestibule.
Threshold and opacity values. Since the statistical phenomena in physical measurements followed the normal distribution, our measurement results must be normally distributed. A goodness-of-fit test was used to examine the normal distribution of both the membranous vestibular CT values and the opacity ones. The results showed the normal distribution. Both normal probability graphs showed nearly linear. The results of these categories of rank, mean, variance, SD, degree of freedom, χ² value, significance, and precise p-value were suitable for normal distribution. The scatter diagram and the linear regression between the membranous vestibular and the opacity values were obtained from Table 4. The linear equation was Y = 0.181X + 20.430 (Y = the percentage opacity value, X = the membranous vestibular CT value, r: 0.771, p < 0.001). Although the opacity should be defined from the membranous CT value, in this study, we used the mean opacity value of 40% for simplicity. The highest threshold value was 160 HU from Table 4.

Postprocessing.
After transferring the image data to a CT workstation (GE Healthcare; Chicago, IL, USA), 3D visualization based on the interactive direct volume rendering was made using GE Advantage Navigator software Ver. 2. Direct volume rendering considers some of the image data, so roughly simple explicit segmentation using a cutting-plane method 36 before the visualization process was required. As a result of this software, both color and opacity values were adjusted interactively to delineate all structures related to the membranous vestibule in real-time. After an appropriate setting using the front cut by a cutting-plane method for optimal delineation of the target structures was defined on axial, coronal, and sagittal projection, the color and opacity table can be stored and used for further studies. The user ensured the intuitive manipulation of any object in real-time. The software allowed both distance measurements and areas directly within the 3D scene. This software was almost the same as the open-source OsiriX software (available at http://www.osirix-viewer.com).
One radiologist specialized in head and neck imaging had 30 years of experience, and one otologist had 40 years of experience.
Quantitative image analysis. Measured dimensions. Utricle. The width of the utricle was measured from the anterior border to the posterior edge, and the length was between the edges of the anterior duct and the common crus in the superior view like Fig. 2.
Saccule. The longitudinal axis was taken to be a line along the surface that seemed to represent the median of the saccule in the frontal view as in Figs. 1 and 3. The width of the saccule was determined by the widest perpendicular line to the longitudinal line.
Lateral semicircular duct. The width of the lateral semicircular duct was measured in the vicinity of the ampulla like Fig. 2.
The angle between utricle and saccule. The angle between them was measured in a 3D lateral view.
Analysis of the age-related changes. We examined the relationships between each measurement value and the age group as follows.