ANATOMICAL STUDY
Normative Inner Ear Volumetric Measurements Michael T. Teixido, MD,*† Gary Kirkilas, DO,‡ Peter Seymour, MD,§ Kanik Sem, MD,|| Alberto Iaia, MD,† Omar Sabra, MD,¶ and Huseyin Isildak, MD# Abstract: In the current study, we attempted to determine normative inner ear volumetric measurements generated from three-dimensional computed tomography (CT) images. In addition, we investigated a correlation between the axial length and the volume of the labyrinth and discussed clinical outcomes of this correlation. Amira 5.2.2 software was used to create three-dimensional isosurface images of the human labyrinth using two-dimensional CT images from 35 anatomically normal patients. With the three-dimensional labyrinths, complete dimensional analysis was performed to gain insight into both the volume and the greatest axial length of the inner ear. Paired t test and Pearson correlation were used. Our volume of the inner ear inquiry reported a mean volume of 221.5 with SD of 24.3 μL (0.228 μL for males and 0.218 μL for females). The length showed a mean of 1.713 cm with SD of 0.064 cm (1.753 cm for males and 1.695 cm for females). The length was used to estimate the volume, and the estimates were within 10% of the measured volume 74.3% of the time. Normative volumetric measurements of the inner ear can be obtained by using threedimensional CT Imaging by Amira 5.2.2 software. There was a statistically significant positive correlation between the axial length of the labyrinth and the volume of the labyrinth. The axial length of the labyrinth could be used to estimate the volume of the labyrinth, which may be clinically important to estimate the concentration of the drug distributed in the inner ear. Key Words: Inner ear volume, the axial length of labyrinth, drug distributed in the inner ear fluids, three-dimensional CT (J Craniofac Surg 2015;26: 251–254)
n 1890, Siebenmann1 performed corrosion casts of the human labyrinth using Wood’s metal and demonstrated anatomy of the human labyrinth in great detail. Inner ear volume has also been
I
From the *Thomas Jefferson University, Philadelphia, Pennsylvania; †Christiana Care Health Systems, Newark, Delaware; ‡Loma Linda Medical Center Loma Linda, California; §Colden & Seymour ENT and Allergy, Newburyport, Massachusetts; ∥University of Delaware, Newark, Delaware; ¶Division of Otolaryngology-Head and Neck Surgery, Saad Specialist Hospital, Al-Khobar, Saudi Arabia; and #Division of Otolaryngology-Head and Neck Surgery, College of Medicine, The Pennsylvania State University, Hershey, Pennsylvania. Received April 15, 2014. Accepted for publication July 1, 2014. O.S. was an Otology Fellow at Christiana Care Health System, Wilmington, Delaware. Address correspondence and reprint requests to Huseyin Isildak, MD, Division of Otolaryngology–Head and Neck Surgery, College of Medicine, The Pennsylvania State University, 500 University Dr, H091, PO Box 850, Hershey, PA 17033; E-mail: [email protected] The authors report no conflicts of interest. Copyright © 2015 by Mutaz B. Habal, MD ISSN: 1049-2275 DOI: 10.1097/SCS.0000000000001204
measured using histological modalities,2 which has more recently been made by using magnetic resonance imaging (MRI).3,4 A semiautomatic segmentation method was used in most of these imaging studies. In this method, the user outlines the region of interest, and algorithms are applied, so that best fit to the edge of the image is shown. Amira 5.2.2 software (Visage Imaging GmbH, Berlin, Germany), which has a semiautomatic segmentation method, is a commercially available software. It allows users to reconstruct threedimensional isosurface structures from complete two-dimensional computed tomography (CT) image stacks. In the current study, we attempted to determine normative inner ear volumetric measurements generated from three-dimensional CT images using Amira 5.2.2 software. In addition, we investigated a correlation between the axial length and the volume of the labyrinth and discussed clinical outcomes of this correlation.
MATERIALS AND METHODS This study was performed solely at the authors’ institution and with institutional review board approval. Reconstructing a three-dimensional labyrinth by way of Amira version 5.2.2 requires a process called segmentation. This process begins with the human observer identifying the structures of the bony labyrinth in a series of contiguous CT images in the Amira viewer panel. The observer identifies structures and their tissue borders based on anatomical definition and visual recognition. Once identification of the structure is made, the user must use Amira’s labeling tools to outline and fill desired anatomical structures with a color of the user’s choice. This process of structural identification and labeling is then repeated in each of the serial CT slices for each patient. The CT data were acquired on 3 types of multidetector CT scanners: a Brilliance 64 (Philips Healthcare, Best, the Netherlands) and Somatom Sensation 64 and Somatom Definition (Siemens Healthcare, Erlangen, Germany). The following parameters for image acquisition were used. For Brilliance 64 and Somatom Sensation, data were acquired at 400mAs, 120 kVp, at 0.6 mm, reconstructed at 0.4-mm increments, with a collimation of 12 0.6 mm, rotation time of 1.0 second, and a pitch of 0.85. Images were reconstructed at 0.8-mm slice thickness, at 0.4-mm intervals. For Somatom Definition, the parameters were the same, with exception of the collimation (16 0.6 mm) and the pitch (0.8). Our serial image stacks comprised approximately 120 contiguous twodimensional CT head images of 0.8-mm thickness for each of the 35 patients’ right ears (24 females, 11 males; age range, 9–86 years;
FIGURE 1. The greatest axial length was calculated from the most posterior portion of the posterior SSC to the most anterior point of the cochlea.
The Journal of Craniofacial Surgery • Volume 26, Number 1, January 2015
Copyright © 2014 Mutaz B. Habal, MD. Unauthorized reproduction of this article is prohibited.
251
The Journal of Craniofacial Surgery • Volume 26, Number 1, January 2015
Teixido et al
TABLE 1. Our Findings Compared Well Against Previous Studies Using MRI and Histological Modalities Volume 227.8 μL, SD of 24.4 μL 221.5 μL SD of 24.3 μL 204.5 μL 192.5 μL
Study Source 3
Modality
Melhem and Shakir, 1998 Current study
High-resolution heavily T2-weighted three-dimensional fast spin-echo magnetic resonance images in 23 subjects 0.8 mm CT slices with Amira 4.1.2 software in 35 subjects
Igarashi et al,2 1986 Buckingham and Valvassori,4 2001
Histological slides with a microcomputer digitizing tablet in 5 subjects Plastic cast made from a single histological sample and measured the amount of water displaced
mean age, 56 years). The CT images from all 35 patients had been previously viewed by an otologist and radiologist. The patients have chronic ear disease; however, they don’t have any structural abnormalities. Labeling each structure of the bony labyrinth was facilitated by a combination of semiautomated labeling tools built into the Amira software followed by manual voxel selection. These tools automatically create structural outlines by differentiating selected structures that radiofluoresce at different shades of black and white (referred to as the voxel range) from the undesired surrounding structures. Manual voxel selection was then performed to exclude unwanted selected volumes such as the internal auditory meatus and to include volume under sampled by the automated method. The user can select both the labeling tool and the voxel range. For our study, we achieved the best results using the “magic wand” tool with a voxel range of −2000 to +995. Thus, for instance, when labeling the vestibule, the magic wand would label the vestibule based on the fact that this specific region contained voxels in the range of −2000 and +995, whereas the surrounding otic capsule fluoresces at voxels greater than +995. In those instances where the bony labyrinth had a similar voxel range as the surrounding bone, the magic wand could not make structural distinctions and therefore could not be used. Instead, the labyrinth was labeled using the “marker” tool, which enables the user to manually select pixels within the −2000 to +995 voxel range. This was useful, for example, when labeling the cochlear turns but not the narrow semicircular canals (SCCs). The marker tool was also used to manually unselect those structures that were incorrectly labeled, for instance, the external and internal auditory canal. The 35 labyrinths were independently segmented by 1 trained observer. Each labyrinth segmentation required 3 hours. For this study, we choose to segment the entire vestibule along with superior, posterior, and horizontal SCCs, whereas the round window, vestibular aqueduct, and endolymphatic sac were not included. Previous volumetric studies also excluded the endolymphatic sac because of inability to view these structures or because there was insufficient fluid in endolymphatic sac to measure. As for the cochlea, the only structures not labeled were the cochlear aqueduct and modiolus. Once the user has completed segmenting the desired structures in each CT slice, the Amira program compiled each segment in the image stack to create the three-dimensional image of the labyrinth. The mechanism by which this takes place is an algorithm referred to as direct volume rendering, where the Amira software can selectively remove nonsegmented optical slices to expose the threedimensional structures that were segmented.5 With the three-dimensional labyrinth complete, dimensional analysis was performed to gain insight on both volume and length. Volume measurements were ascertained from using Amira’s TissueStatistics function. This function automatically computes the total internal volume of the labyrinth. Length measurements were computed by first positioning our labyrinths in a standard three-dimensional orientation. Amira allows the user to manipulate the three-dimensional orientation of
252
structure in the viewer panel. Labyrinths were positioned horizontally from a superior view so that the superior and inferior crus of posterior canals overlaid one another. We then adjusted the pitch of labyrinth to the point that the posterior surface of the superior SCC contacted the anterior lumen of the horizontal SCC. Then, using the two-dimensional ruler function, the longest length dimension could be calculated. In this standardized position, the greatest
TABLE 2. Our Volume and Length Results Patient 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 Mean Min Max SD Male mean Female mean
Sex
Volume, μL
Length, cm
Female Female Male Female Male Female Male Female Female Female Female Male Male Female Female Female Female Male Female Female Female Male Female Male Female Female Female Female Male Female Male Female Female Male Female
0.20553 0.188423 0.203988 0.226002 0.235521 0.24744 0.297501 0.213905 0.227231 0.226573 0.263628 0.248142 0.21263 0.211365 0.239449 0.229003 0.21934 0.220649 0.208063 0.210403 0.200693 0.223758 0.24844 0.20732 0.189164 0.188647 0.203419 0.267196 0.198936 0.192271 0.244081 0.207842 0.203992 0.219461 0.221504 0.221472 0.188423 0.297501 0.024294 0.228 0.218
1.74 1.61 1.76 1.72 1.73 1.61 1.84 1.66 1.7 1.65 1.77 1.77 1.71 1.72 1.76 1.78 1.72 1.74 1.65 1.69 1.66 1.82 1.69 1.66 1.64 1.67 1.69 1.88 1.72 1.66 1.79 1.63 1.66 1.75 1.72 1.713429 1.61 1.88 0.063983 1.753 1.695
© 2015 Mutaz B. Habal, MD
Copyright © 2014 Mutaz B. Habal, MD. Unauthorized reproduction of this article is prohibited.
The Journal of Craniofacial Surgery • Volume 26, Number 1, January 2015
FIGURE 2. A, The axial length of the labyrinth was statistically significant longer in males. B, The nonsignificant difference between males and females in the volume of the labyrinth.
length was calculated from the most posterior portion of the posterior SSC to the most anterior point of the cochlea (Fig. 1).
RESULTS AND ANALYSIS
Our volume inquiry reported a mean volume of 221.5 μL, with a minimum volume of 188.3 μL, a maximum volume of 297.4 μL, and an SD of 24.3 μL. Volume compared with gender showed a mean volume of 0.228 μL for males and 0.218 μL for females. Our volume results are summarized in Table 1. Our findings compared well against previous studies using MRI and histological modalities (Table 1). The greatest length showed a mean of 1.713 cm, a minimum length of 1.61, a maximum length of 1.88 cm, and an SD of 0.064 cm. Length compared with gender showed a mean of 1.753 cm for males and 1.695 cm for females (Table 2). Although there was no statistically significant difference between males and females in the volume of the labyrinth (P = 0.26), the axial length of the labyrinth was statistically significant longer in males (P = 0.01) (Figs. 2A, B). We applied the t test on the data values and found that there is a statistically significant positive correlation between the axial length and the volume of the labyrinth (r = 0.645 P < 0.01) (Fig. 3). The volume can be estimated as follows: volume = −0.119191 + 0.187092 L, where L is the greatest axial length calculated from the most posterior portion of the posterior SSC to the most anterior point of the cochlea. Percent error and absolute percent error scores were calculated for each estimation result. Mean absolute percent error was 7.068% (SD, 6.00%) (Figs. 4A, B). The estimates were within 10% of the measured volume 74.3% of the time.
DISCUSSION In the current study, CT was done for chronic ear disease; however, the absence of any gross structural abnormalities was confirmed. Only right ears were assessed. Although there are limited studies with which to compare our data, we did find that our results were similar to published data using a variety of methodological approaches. In 1986, Igarashi and colleagues2 used histological slides from 5 human temporal bones and computed the total volume of the bony labyrinth to be 204.5 μL using a microcomputer digitizing tablet. This is a difference of only 17 μL from our value. In 2001, Buckingham and Valvassori4 completed a study using a plastic cast from a histological sample from a single human temporal bone. They computed a volume of 192.5 μL by measuring the amount of water displaced by submerging the plastic cast.4 Closest in methodology to our study was Melhem and Shakir’s3 1998 study, which used high-resolution heavily T2-weighted three-dimensional fast spin-echo magnetic resonance images. They computed a total volume of 227.8 μL from 23 volunteers with an SD of 24.4 μL. This
Normative Inner Ear Volumetric Measurements
value was closest to ours with a difference of 6.3 μL in total volume and a difference of 0.1 μL in SD. The choice of modalities to assess volumes is fraught with bias and human error. One explanation for the fact that histologically prepared volume inquiries consistently report smaller values is that a shrinkage artifact exists in the preparation process of human temporal bones. Buckingham and Valvassori referenced one previous study that suggested shrinkage of approximately 10%.4 Another inconsistency with volume studies that use histological slides is tissue distortion in the slide placement process. It is known that some tissue distortion takes places when placing the histological specimens on the microscope slides causing volume inquiries to vary depending on the amount of distortion. In regard to our study, it was evident that in using Amira’s labeling tools a significant degree human observer bias was present in subjective recognition of tissue boarders. To eliminate some of the bias, a standard voxel range was used (−2000 to +995) in selecting structures. This was useful in segmenting structures with large voxel range differences, for example, the vestibule and the surrounding otic capsule. However, when structures had similar voxel ranges, it required the observer to subjectively differentiate tissue borders. For example, segmenting the cochlea required a subjective decision on the part of the human observer to differentiate between the desired cochlea and undesired modiolus, both structures with similar voxel ranges. Perforations in the modiolus described by Plontke et al6 may change volumes slightly as the volume of Rosenthal’s canal has not in this study, nor has it traditionally been included in volume measurements. There is no completely automated method for rendering the volume of structures as small as the bony labyrinth at clinical CT resolution. In the current study, we used conventional CT with 0.8-mm slice thickness. Clinical CT scan also delivers 0.65- and 0.5-mm slice data at acceptable radiation doses. Peltonen et al7 reported that the effective dose for cone-beam CT imaging of the middle ear is 13 μSv, which is 60 times lower than that of a conventional multidetector CT scan of the temporal bone. Cone-beam CT may offer exceptionally accurate measurements as slice thickness of 0.1 mm can be achieved with acceptable radiation dose to the patient. We believe that similar research to ours may be executed by using cone-beam CT in the future and that the need for subjective voxel labeling may be substantially eliminated. When using CT, MRI, and electron microscopy in volume studies, much debate exists on which structures should be included in the segmentation process. It was our reasoning that accurately segmenting the endolymphatic sac and duct was not feasible without bony boundaries. Also, our decision to exclude the bony modiolus could be argued against because of the anatomical fenestrations that exist within the structure that communicate the perilymphatic space and Rosenthal's canal.
FIGURE 3. The positive correlation between the axial length and the volume of the labyrinth.
© 2015 Mutaz B. Habal, MD
Copyright © 2014 Mutaz B. Habal, MD. Unauthorized reproduction of this article is prohibited.
253
Teixido et al
The Journal of Craniofacial Surgery • Volume 26, Number 1, January 2015
FIGURE 4. Percent error (A) and absolute percent error (B) for each estimation result.
Studies show that inner ear fluids do not circulate. Therefore, the topical drugs diffuse to the inner ear slowly, mainly by passive diffusion. Although the rate of clearance of the drug from the inner ear is the major factor that determines the drug concentration in the inner ear, the fluid volume in the inner ear cannot be ignored. Inner ear volume may be clinically important to estimate the concentration of a drug distributed in the inner ear fluids after local application of medications. Transtympanic drug administration is common in otolaryngology practice. The optimal dosage and dosing intervals are highly variable among clinicians and clinical outcomes of treatment vary as well. As an example, whereas some authors have reported complete deafness in more than 20%8 or even 80%,9 others have reported no hearing loss following single or repeated transtympanic administration of gentamicin for Ménière disease. A detailed discussion of this evolving field is beyond the scope of this report. There are number of studies that use a computer simulation program that calculates drug dispersal in the inner ear in animals6,10 and in humans.11 However they could calculate drug dispersal only in the cochlear fluids. They excluded the vestibular system, because of the complex geometry of the vestibular spaces. Salt et al11 report that higher doses of gentamicin in the cochlear fluids damage hearing in some cases. Accurate inner ear volume data may be important for calculation of drug dose and prevention of hearing loss. Some authors claim that inner ear volumetric measurements may help identify patients with congenital sensorineural hearing loss and normal inner ear configuration.3 The association between normal configuration but abnormal size and hearing loss or balance disorders deserves investigation. Our findings for labyrinth length are difficult to compare with other studies as our method of measuring was unique to our study. In addition, there are no studies that specifically measured the greatest length of the bony labyrinth. All previous studies measured specific regions of the labyrinth (ie, diameter of the SCCs). In our study, we also showed a correlation between the axial length and the volume of the labyrinth. The length is not the length of the conical cochlear duct, but is the overall length of the entire complex structure of the labyrinth, which can vary in its width, height, and degree of coiling. Therefore, we expected to find much larger variations in the estimated volumes. In our study, however, the estimates were within 10% of the measured volume 74.3% of the time. It implies that the length is a predictor of the volume of the labyrinth. Because the axial length of the labyrinth is easy to measure clinically, we suggest that the length could be used to estimate the volume of the labyrinth and thus could inform estimates of appropriate drug dose. The complex structure of the inner ear may show variations in gender. Although the axial length of the labyrinth was statistically significantly longer in males (P = 0.01),
254
there was no statistically significant difference between males and females in the volume of the labyrinth (P = 0.26). In conclusion, normative volumetric measurements of the inner ear can be obtained by using three-dimensional CT imaging by Amira software. The axial length of the labyrinth was statistically significantly longer in males, whereas the volume is about the same in both groups. The axial length of the labyrinth could be used to estimate the volume of the labyrinth. This relationship may become clinically important to the future of drug delivery and estimations of the concentration of drug distributed in the inner ear. ACKNOWLEDGMENT The authors thank Alec Salt, PhD, for his comments regarding the preparation of these data.
REFERENCES 1. Siebenmann F. Die Korrosions-Anatomie des knochernen Labyrinthes des menschlichen Ohres. Wiesbaden, Germany: C. F. Bergmann, 1890 2. Igarashi M, Ohashi K, Ishii M. Morphometric comparison of endolymphatic and perilymphatic spaces in human temporal bones. Acta Otolaryngol 1986;101:161–164 3. Melhem ER, Shakir H, Bakthavachalam S, et al. Inner ear volumetric measurements using high-resolution 3D T2-weighted fast spin-echo MR imaging: initial experience in healthy subjects. AJNR Am J Neuroradiol 1998;19:1819–1822 4. Buckingham RA, Valvassori GE. Inner ear fluid volumes and the resolving power of magnetic resonance imaging: can it differentiate endolymphatic structures? Ann Otol Rhinol Laryngol 2001;110:113–117 5. Santi PA, Rapson I, Voie A. Development of the mouse cochlea database (MCD). Hear Res 2008;243:11–17 6. Plontke SK, Wood AW, Salt AN. Analysis of gentamicin kinetics in fluids of the inner ear with round window administration. Otol Neurotol 2002;23:967–974 7. Peltonen LI, Aarnisalo AA, Kortesniemi MK, et al. Limited cone-beam computed tomography imaging of the middle ear: a comparison with multislice helical computed tomography. Acta Radiol 2007;48:207–212 8. Thomsen J, Charabi S, Tos M. Preliminary results of a new delivery system for gentamicin to the inner ear in patients with Ménière's disease. Eur Arch Otorhinolaryngol 2000;257:362–365 9. Schoendorf J, Neugebauer P, Michel O. Continuous intratympanic infusion of gentamicin via a microcatheter in Ménière's disease. Otolaryngol Head Neck Surg 2001;124:203–207 10. Mynatt R, Hale SA, Gill RM, et al. Demonstration of a longitudinal concentration gradient along scala tympani by sequential sampling of perilymph from the cochlear apex. J Assoc Res Otolaryngol 2006;7:182–193 11. Salt AN, Gill RM, Plontke SK. Dependence of hearing changes on the dose of intratympanically applied gentamicin: a meta-analysis using mathematical simulations of clinical drug delivery protocols. Laryngoscope 2008;118:1793–1800
© 2015 Mutaz B. Habal, MD
Copyright © 2014 Mutaz B. Habal, MD. Unauthorized reproduction of this article is prohibited.
Normative Inner Ear Volumetric Measurements Michael T. Teixido, MD,*† Gary Kirkilas, DO,‡ Peter Seymour, MD,§ Kanik Sem, MD,|| Alberto Iaia, MD,† Omar Sabra, MD,¶ and Huseyin Isildak, MD# Abstract: In the current study, we attempted to determine normative inner ear volumetric measurements generated from three-dimensional computed tomography (CT) images. In addition, we investigated a correlation between the axial length and the volume of the labyrinth and discussed clinical outcomes of this correlation. Amira 5.2.2 software was used to create three-dimensional isosurface images of the human labyrinth using two-dimensional CT images from 35 anatomically normal patients. With the three-dimensional labyrinths, complete dimensional analysis was performed to gain insight into both the volume and the greatest axial length of the inner ear. Paired t test and Pearson correlation were used. Our volume of the inner ear inquiry reported a mean volume of 221.5 with SD of 24.3 μL (0.228 μL for males and 0.218 μL for females). The length showed a mean of 1.713 cm with SD of 0.064 cm (1.753 cm for males and 1.695 cm for females). The length was used to estimate the volume, and the estimates were within 10% of the measured volume 74.3% of the time. Normative volumetric measurements of the inner ear can be obtained by using threedimensional CT Imaging by Amira 5.2.2 software. There was a statistically significant positive correlation between the axial length of the labyrinth and the volume of the labyrinth. The axial length of the labyrinth could be used to estimate the volume of the labyrinth, which may be clinically important to estimate the concentration of the drug distributed in the inner ear. Key Words: Inner ear volume, the axial length of labyrinth, drug distributed in the inner ear fluids, three-dimensional CT (J Craniofac Surg 2015;26: 251–254)
n 1890, Siebenmann1 performed corrosion casts of the human labyrinth using Wood’s metal and demonstrated anatomy of the human labyrinth in great detail. Inner ear volume has also been
I
From the *Thomas Jefferson University, Philadelphia, Pennsylvania; †Christiana Care Health Systems, Newark, Delaware; ‡Loma Linda Medical Center Loma Linda, California; §Colden & Seymour ENT and Allergy, Newburyport, Massachusetts; ∥University of Delaware, Newark, Delaware; ¶Division of Otolaryngology-Head and Neck Surgery, Saad Specialist Hospital, Al-Khobar, Saudi Arabia; and #Division of Otolaryngology-Head and Neck Surgery, College of Medicine, The Pennsylvania State University, Hershey, Pennsylvania. Received April 15, 2014. Accepted for publication July 1, 2014. O.S. was an Otology Fellow at Christiana Care Health System, Wilmington, Delaware. Address correspondence and reprint requests to Huseyin Isildak, MD, Division of Otolaryngology–Head and Neck Surgery, College of Medicine, The Pennsylvania State University, 500 University Dr, H091, PO Box 850, Hershey, PA 17033; E-mail: [email protected] The authors report no conflicts of interest. Copyright © 2015 by Mutaz B. Habal, MD ISSN: 1049-2275 DOI: 10.1097/SCS.0000000000001204
measured using histological modalities,2 which has more recently been made by using magnetic resonance imaging (MRI).3,4 A semiautomatic segmentation method was used in most of these imaging studies. In this method, the user outlines the region of interest, and algorithms are applied, so that best fit to the edge of the image is shown. Amira 5.2.2 software (Visage Imaging GmbH, Berlin, Germany), which has a semiautomatic segmentation method, is a commercially available software. It allows users to reconstruct threedimensional isosurface structures from complete two-dimensional computed tomography (CT) image stacks. In the current study, we attempted to determine normative inner ear volumetric measurements generated from three-dimensional CT images using Amira 5.2.2 software. In addition, we investigated a correlation between the axial length and the volume of the labyrinth and discussed clinical outcomes of this correlation.
MATERIALS AND METHODS This study was performed solely at the authors’ institution and with institutional review board approval. Reconstructing a three-dimensional labyrinth by way of Amira version 5.2.2 requires a process called segmentation. This process begins with the human observer identifying the structures of the bony labyrinth in a series of contiguous CT images in the Amira viewer panel. The observer identifies structures and their tissue borders based on anatomical definition and visual recognition. Once identification of the structure is made, the user must use Amira’s labeling tools to outline and fill desired anatomical structures with a color of the user’s choice. This process of structural identification and labeling is then repeated in each of the serial CT slices for each patient. The CT data were acquired on 3 types of multidetector CT scanners: a Brilliance 64 (Philips Healthcare, Best, the Netherlands) and Somatom Sensation 64 and Somatom Definition (Siemens Healthcare, Erlangen, Germany). The following parameters for image acquisition were used. For Brilliance 64 and Somatom Sensation, data were acquired at 400mAs, 120 kVp, at 0.6 mm, reconstructed at 0.4-mm increments, with a collimation of 12 0.6 mm, rotation time of 1.0 second, and a pitch of 0.85. Images were reconstructed at 0.8-mm slice thickness, at 0.4-mm intervals. For Somatom Definition, the parameters were the same, with exception of the collimation (16 0.6 mm) and the pitch (0.8). Our serial image stacks comprised approximately 120 contiguous twodimensional CT head images of 0.8-mm thickness for each of the 35 patients’ right ears (24 females, 11 males; age range, 9–86 years;
FIGURE 1. The greatest axial length was calculated from the most posterior portion of the posterior SSC to the most anterior point of the cochlea.
The Journal of Craniofacial Surgery • Volume 26, Number 1, January 2015
Copyright © 2014 Mutaz B. Habal, MD. Unauthorized reproduction of this article is prohibited.
251
The Journal of Craniofacial Surgery • Volume 26, Number 1, January 2015
Teixido et al
TABLE 1. Our Findings Compared Well Against Previous Studies Using MRI and Histological Modalities Volume 227.8 μL, SD of 24.4 μL 221.5 μL SD of 24.3 μL 204.5 μL 192.5 μL
Study Source 3
Modality
Melhem and Shakir, 1998 Current study
High-resolution heavily T2-weighted three-dimensional fast spin-echo magnetic resonance images in 23 subjects 0.8 mm CT slices with Amira 4.1.2 software in 35 subjects
Igarashi et al,2 1986 Buckingham and Valvassori,4 2001
Histological slides with a microcomputer digitizing tablet in 5 subjects Plastic cast made from a single histological sample and measured the amount of water displaced
mean age, 56 years). The CT images from all 35 patients had been previously viewed by an otologist and radiologist. The patients have chronic ear disease; however, they don’t have any structural abnormalities. Labeling each structure of the bony labyrinth was facilitated by a combination of semiautomated labeling tools built into the Amira software followed by manual voxel selection. These tools automatically create structural outlines by differentiating selected structures that radiofluoresce at different shades of black and white (referred to as the voxel range) from the undesired surrounding structures. Manual voxel selection was then performed to exclude unwanted selected volumes such as the internal auditory meatus and to include volume under sampled by the automated method. The user can select both the labeling tool and the voxel range. For our study, we achieved the best results using the “magic wand” tool with a voxel range of −2000 to +995. Thus, for instance, when labeling the vestibule, the magic wand would label the vestibule based on the fact that this specific region contained voxels in the range of −2000 and +995, whereas the surrounding otic capsule fluoresces at voxels greater than +995. In those instances where the bony labyrinth had a similar voxel range as the surrounding bone, the magic wand could not make structural distinctions and therefore could not be used. Instead, the labyrinth was labeled using the “marker” tool, which enables the user to manually select pixels within the −2000 to +995 voxel range. This was useful, for example, when labeling the cochlear turns but not the narrow semicircular canals (SCCs). The marker tool was also used to manually unselect those structures that were incorrectly labeled, for instance, the external and internal auditory canal. The 35 labyrinths were independently segmented by 1 trained observer. Each labyrinth segmentation required 3 hours. For this study, we choose to segment the entire vestibule along with superior, posterior, and horizontal SCCs, whereas the round window, vestibular aqueduct, and endolymphatic sac were not included. Previous volumetric studies also excluded the endolymphatic sac because of inability to view these structures or because there was insufficient fluid in endolymphatic sac to measure. As for the cochlea, the only structures not labeled were the cochlear aqueduct and modiolus. Once the user has completed segmenting the desired structures in each CT slice, the Amira program compiled each segment in the image stack to create the three-dimensional image of the labyrinth. The mechanism by which this takes place is an algorithm referred to as direct volume rendering, where the Amira software can selectively remove nonsegmented optical slices to expose the threedimensional structures that were segmented.5 With the three-dimensional labyrinth complete, dimensional analysis was performed to gain insight on both volume and length. Volume measurements were ascertained from using Amira’s TissueStatistics function. This function automatically computes the total internal volume of the labyrinth. Length measurements were computed by first positioning our labyrinths in a standard three-dimensional orientation. Amira allows the user to manipulate the three-dimensional orientation of
252
structure in the viewer panel. Labyrinths were positioned horizontally from a superior view so that the superior and inferior crus of posterior canals overlaid one another. We then adjusted the pitch of labyrinth to the point that the posterior surface of the superior SCC contacted the anterior lumen of the horizontal SCC. Then, using the two-dimensional ruler function, the longest length dimension could be calculated. In this standardized position, the greatest
TABLE 2. Our Volume and Length Results Patient 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 Mean Min Max SD Male mean Female mean
Sex
Volume, μL
Length, cm
Female Female Male Female Male Female Male Female Female Female Female Male Male Female Female Female Female Male Female Female Female Male Female Male Female Female Female Female Male Female Male Female Female Male Female
0.20553 0.188423 0.203988 0.226002 0.235521 0.24744 0.297501 0.213905 0.227231 0.226573 0.263628 0.248142 0.21263 0.211365 0.239449 0.229003 0.21934 0.220649 0.208063 0.210403 0.200693 0.223758 0.24844 0.20732 0.189164 0.188647 0.203419 0.267196 0.198936 0.192271 0.244081 0.207842 0.203992 0.219461 0.221504 0.221472 0.188423 0.297501 0.024294 0.228 0.218
1.74 1.61 1.76 1.72 1.73 1.61 1.84 1.66 1.7 1.65 1.77 1.77 1.71 1.72 1.76 1.78 1.72 1.74 1.65 1.69 1.66 1.82 1.69 1.66 1.64 1.67 1.69 1.88 1.72 1.66 1.79 1.63 1.66 1.75 1.72 1.713429 1.61 1.88 0.063983 1.753 1.695
© 2015 Mutaz B. Habal, MD
Copyright © 2014 Mutaz B. Habal, MD. Unauthorized reproduction of this article is prohibited.
The Journal of Craniofacial Surgery • Volume 26, Number 1, January 2015
FIGURE 2. A, The axial length of the labyrinth was statistically significant longer in males. B, The nonsignificant difference between males and females in the volume of the labyrinth.
length was calculated from the most posterior portion of the posterior SSC to the most anterior point of the cochlea (Fig. 1).
RESULTS AND ANALYSIS
Our volume inquiry reported a mean volume of 221.5 μL, with a minimum volume of 188.3 μL, a maximum volume of 297.4 μL, and an SD of 24.3 μL. Volume compared with gender showed a mean volume of 0.228 μL for males and 0.218 μL for females. Our volume results are summarized in Table 1. Our findings compared well against previous studies using MRI and histological modalities (Table 1). The greatest length showed a mean of 1.713 cm, a minimum length of 1.61, a maximum length of 1.88 cm, and an SD of 0.064 cm. Length compared with gender showed a mean of 1.753 cm for males and 1.695 cm for females (Table 2). Although there was no statistically significant difference between males and females in the volume of the labyrinth (P = 0.26), the axial length of the labyrinth was statistically significant longer in males (P = 0.01) (Figs. 2A, B). We applied the t test on the data values and found that there is a statistically significant positive correlation between the axial length and the volume of the labyrinth (r = 0.645 P < 0.01) (Fig. 3). The volume can be estimated as follows: volume = −0.119191 + 0.187092 L, where L is the greatest axial length calculated from the most posterior portion of the posterior SSC to the most anterior point of the cochlea. Percent error and absolute percent error scores were calculated for each estimation result. Mean absolute percent error was 7.068% (SD, 6.00%) (Figs. 4A, B). The estimates were within 10% of the measured volume 74.3% of the time.
DISCUSSION In the current study, CT was done for chronic ear disease; however, the absence of any gross structural abnormalities was confirmed. Only right ears were assessed. Although there are limited studies with which to compare our data, we did find that our results were similar to published data using a variety of methodological approaches. In 1986, Igarashi and colleagues2 used histological slides from 5 human temporal bones and computed the total volume of the bony labyrinth to be 204.5 μL using a microcomputer digitizing tablet. This is a difference of only 17 μL from our value. In 2001, Buckingham and Valvassori4 completed a study using a plastic cast from a histological sample from a single human temporal bone. They computed a volume of 192.5 μL by measuring the amount of water displaced by submerging the plastic cast.4 Closest in methodology to our study was Melhem and Shakir’s3 1998 study, which used high-resolution heavily T2-weighted three-dimensional fast spin-echo magnetic resonance images. They computed a total volume of 227.8 μL from 23 volunteers with an SD of 24.4 μL. This
Normative Inner Ear Volumetric Measurements
value was closest to ours with a difference of 6.3 μL in total volume and a difference of 0.1 μL in SD. The choice of modalities to assess volumes is fraught with bias and human error. One explanation for the fact that histologically prepared volume inquiries consistently report smaller values is that a shrinkage artifact exists in the preparation process of human temporal bones. Buckingham and Valvassori referenced one previous study that suggested shrinkage of approximately 10%.4 Another inconsistency with volume studies that use histological slides is tissue distortion in the slide placement process. It is known that some tissue distortion takes places when placing the histological specimens on the microscope slides causing volume inquiries to vary depending on the amount of distortion. In regard to our study, it was evident that in using Amira’s labeling tools a significant degree human observer bias was present in subjective recognition of tissue boarders. To eliminate some of the bias, a standard voxel range was used (−2000 to +995) in selecting structures. This was useful in segmenting structures with large voxel range differences, for example, the vestibule and the surrounding otic capsule. However, when structures had similar voxel ranges, it required the observer to subjectively differentiate tissue borders. For example, segmenting the cochlea required a subjective decision on the part of the human observer to differentiate between the desired cochlea and undesired modiolus, both structures with similar voxel ranges. Perforations in the modiolus described by Plontke et al6 may change volumes slightly as the volume of Rosenthal’s canal has not in this study, nor has it traditionally been included in volume measurements. There is no completely automated method for rendering the volume of structures as small as the bony labyrinth at clinical CT resolution. In the current study, we used conventional CT with 0.8-mm slice thickness. Clinical CT scan also delivers 0.65- and 0.5-mm slice data at acceptable radiation doses. Peltonen et al7 reported that the effective dose for cone-beam CT imaging of the middle ear is 13 μSv, which is 60 times lower than that of a conventional multidetector CT scan of the temporal bone. Cone-beam CT may offer exceptionally accurate measurements as slice thickness of 0.1 mm can be achieved with acceptable radiation dose to the patient. We believe that similar research to ours may be executed by using cone-beam CT in the future and that the need for subjective voxel labeling may be substantially eliminated. When using CT, MRI, and electron microscopy in volume studies, much debate exists on which structures should be included in the segmentation process. It was our reasoning that accurately segmenting the endolymphatic sac and duct was not feasible without bony boundaries. Also, our decision to exclude the bony modiolus could be argued against because of the anatomical fenestrations that exist within the structure that communicate the perilymphatic space and Rosenthal's canal.
FIGURE 3. The positive correlation between the axial length and the volume of the labyrinth.
© 2015 Mutaz B. Habal, MD
Copyright © 2014 Mutaz B. Habal, MD. Unauthorized reproduction of this article is prohibited.
253
Teixido et al
The Journal of Craniofacial Surgery • Volume 26, Number 1, January 2015
FIGURE 4. Percent error (A) and absolute percent error (B) for each estimation result.
Studies show that inner ear fluids do not circulate. Therefore, the topical drugs diffuse to the inner ear slowly, mainly by passive diffusion. Although the rate of clearance of the drug from the inner ear is the major factor that determines the drug concentration in the inner ear, the fluid volume in the inner ear cannot be ignored. Inner ear volume may be clinically important to estimate the concentration of a drug distributed in the inner ear fluids after local application of medications. Transtympanic drug administration is common in otolaryngology practice. The optimal dosage and dosing intervals are highly variable among clinicians and clinical outcomes of treatment vary as well. As an example, whereas some authors have reported complete deafness in more than 20%8 or even 80%,9 others have reported no hearing loss following single or repeated transtympanic administration of gentamicin for Ménière disease. A detailed discussion of this evolving field is beyond the scope of this report. There are number of studies that use a computer simulation program that calculates drug dispersal in the inner ear in animals6,10 and in humans.11 However they could calculate drug dispersal only in the cochlear fluids. They excluded the vestibular system, because of the complex geometry of the vestibular spaces. Salt et al11 report that higher doses of gentamicin in the cochlear fluids damage hearing in some cases. Accurate inner ear volume data may be important for calculation of drug dose and prevention of hearing loss. Some authors claim that inner ear volumetric measurements may help identify patients with congenital sensorineural hearing loss and normal inner ear configuration.3 The association between normal configuration but abnormal size and hearing loss or balance disorders deserves investigation. Our findings for labyrinth length are difficult to compare with other studies as our method of measuring was unique to our study. In addition, there are no studies that specifically measured the greatest length of the bony labyrinth. All previous studies measured specific regions of the labyrinth (ie, diameter of the SCCs). In our study, we also showed a correlation between the axial length and the volume of the labyrinth. The length is not the length of the conical cochlear duct, but is the overall length of the entire complex structure of the labyrinth, which can vary in its width, height, and degree of coiling. Therefore, we expected to find much larger variations in the estimated volumes. In our study, however, the estimates were within 10% of the measured volume 74.3% of the time. It implies that the length is a predictor of the volume of the labyrinth. Because the axial length of the labyrinth is easy to measure clinically, we suggest that the length could be used to estimate the volume of the labyrinth and thus could inform estimates of appropriate drug dose. The complex structure of the inner ear may show variations in gender. Although the axial length of the labyrinth was statistically significantly longer in males (P = 0.01),
254
there was no statistically significant difference between males and females in the volume of the labyrinth (P = 0.26). In conclusion, normative volumetric measurements of the inner ear can be obtained by using three-dimensional CT imaging by Amira software. The axial length of the labyrinth was statistically significantly longer in males, whereas the volume is about the same in both groups. The axial length of the labyrinth could be used to estimate the volume of the labyrinth. This relationship may become clinically important to the future of drug delivery and estimations of the concentration of drug distributed in the inner ear. ACKNOWLEDGMENT The authors thank Alec Salt, PhD, for his comments regarding the preparation of these data.
REFERENCES 1. Siebenmann F. Die Korrosions-Anatomie des knochernen Labyrinthes des menschlichen Ohres. Wiesbaden, Germany: C. F. Bergmann, 1890 2. Igarashi M, Ohashi K, Ishii M. Morphometric comparison of endolymphatic and perilymphatic spaces in human temporal bones. Acta Otolaryngol 1986;101:161–164 3. Melhem ER, Shakir H, Bakthavachalam S, et al. Inner ear volumetric measurements using high-resolution 3D T2-weighted fast spin-echo MR imaging: initial experience in healthy subjects. AJNR Am J Neuroradiol 1998;19:1819–1822 4. Buckingham RA, Valvassori GE. Inner ear fluid volumes and the resolving power of magnetic resonance imaging: can it differentiate endolymphatic structures? Ann Otol Rhinol Laryngol 2001;110:113–117 5. Santi PA, Rapson I, Voie A. Development of the mouse cochlea database (MCD). Hear Res 2008;243:11–17 6. Plontke SK, Wood AW, Salt AN. Analysis of gentamicin kinetics in fluids of the inner ear with round window administration. Otol Neurotol 2002;23:967–974 7. Peltonen LI, Aarnisalo AA, Kortesniemi MK, et al. Limited cone-beam computed tomography imaging of the middle ear: a comparison with multislice helical computed tomography. Acta Radiol 2007;48:207–212 8. Thomsen J, Charabi S, Tos M. Preliminary results of a new delivery system for gentamicin to the inner ear in patients with Ménière's disease. Eur Arch Otorhinolaryngol 2000;257:362–365 9. Schoendorf J, Neugebauer P, Michel O. Continuous intratympanic infusion of gentamicin via a microcatheter in Ménière's disease. Otolaryngol Head Neck Surg 2001;124:203–207 10. Mynatt R, Hale SA, Gill RM, et al. Demonstration of a longitudinal concentration gradient along scala tympani by sequential sampling of perilymph from the cochlear apex. J Assoc Res Otolaryngol 2006;7:182–193 11. Salt AN, Gill RM, Plontke SK. Dependence of hearing changes on the dose of intratympanically applied gentamicin: a meta-analysis using mathematical simulations of clinical drug delivery protocols. Laryngoscope 2008;118:1793–1800
© 2015 Mutaz B. Habal, MD
Copyright © 2014 Mutaz B. Habal, MD. Unauthorized reproduction of this article is prohibited.
