Posterior Pole Asymmetry Analyses of Retinal Thickness of Upper and Lower Sectors and Their Association With Peak Retinal Nerve Fiber Layer Thickness in Healthy Young Eyes

Abstract

Purpose.: To determine the symmetry of the retinal thicknesses (RT) between the 32 pairs of superior and inferior sectors by posterior pole asymmetrical analysis (PPAA) of the spectral-domain optical coherence tomographic (SD-OCT) images in healthy eyes. In addition, to determine their association with the position of the peak retinal nerve fiber layer (RNFL) thickness.

Methods.: A prospective, observational, cross-sectional study of 64 right eyes. The Spectralis SD-OCT was used to obtain the images, and the PPAA determined the RT of the 64 cells within the central 24° area. The program also compared the thicknesses of corresponding cells across the fovea–disc axis. Circular scans were used to measure the supra- and infratemporal RNFL peak angle differences (PADs). The relationships between the RT of the corresponding cells and the relationship between the differences of the RT of the corresponding cells and PAD were investigated by linear regression analysis.

Results.: The mean differences between the RT of corresponding cells ranged from 3.1 to 23.2 μm. The RT of all upper cells were significantly correlated with the RT of the corresponding lower cells (R = 0.45–0.97, P < 0.001). The coefficients of correlation between the corresponding pairs of central- and temporal-macular cells were higher than that of the peripheral and nasal-macular cells. The differences of the pairs of nasal-macular cells RT were significantly correlated with the PAD.

Conclusions.: The symmetry of the RT between the upper and lower cells was high in the central and temporal-macular areas but not in the peripheral and nasal-macular areas. (www.umin.ac.jp/ctr number, UMIN000006040.)

Introduction

Evidence has been accumulating that the retinal thickness (RT) determined by optical coherence tomography (OCT) ^1–3 is an important parameter for diagnosing different types of retinal diseases. ^4–6 Recently, an asymmetry test, called posterior pole asymmetry analysis (PPAA) by OCT, was introduced for diagnosing glaucoma and is expected to improve the accuracy of diagnosing glaucoma. ⁶ The PPAA, installed in the Spectralis-OCT, can calculate the RT of 64 (8 × 8) cells within the central 24° area and can also perform comparisons of the corresponding cells across the fovea–disc axis. Its diagnostic accuracy was reported to be similar to that obtained by retinal nerve fiber layer (RNFL) thickness circle scans. ^7–10  

The PPAA is based on the hypothesis that the hemispherical symmetry of the RT is altered by an asymmetrical thinning of the RNFL, ganglion cell layer, and inner plexiform layer in glaucomatous eyes. ⁶ However, to the best of our knowledge, there is no detailed report on the symmetry of the retinal thicknesses in healthy subjects especially in the data obtained by spectral-domain (SD)-OCT PPAA. An earlier study showed that a significant correlation was present between the RNFL thicknesses of the supra- to infratemporal cells across the fovea–disc axis. ¹¹ However, the coefficient of determination was only 0.112, a correlation coefficient of 0.335. ¹¹ In addition, there were some healthy eyes with asymmetry of the thick retinal nerve fiber bundles in the superior and inferior sectors. Thus, this asymmetry of the RNFL thickness may not be rare or there may be a factor that alters the hemispherical symmetry of the RT. 

To improve the diagnostic accuracy of PPAA, it is necessary to know the actual symmetry of healthy eyes and the factors affecting the measurements. Thus, we first investigated the relationship between corresponding cells across the fovea–disc axis. Then, with the hypothesis that the deviation of RNFL peak affects the hemispherical symmetry, the correlation of the RNFL peak, and the RT symmetry of healthy eyes was evaluated. 

Methods

All of the procedures used conformed to the tenets of the Declaration of Helsinki. Written, informed consent was obtained from all of the subjects after an explanation of the procedures to be used. The study was approved by the ethics committee of Kagoshima University Hospital (Kagoshima, Japan), and it was registered with the University Hospital Medical Network (UMIN)-clinical trials registry (CTR). The registration title was, “Morphological analysis of the optic disc and the retinal nerve fiber in myopic eyes” and the registration number was UMIN000006040. A detailed protocol is available (in the public domain) at https://upload.umin.ac.jp/cgi-open-bin/ctr/ctr.cgi?function=brows&action=brows&type=summary&recptno=R000007154&language=J. The results presented in this manuscript are part of the overall study. 

Subjects

This was a cross-sectional, prospective, observational study of 72 eyes of 72 volunteers that were examined between February 1, 2011 and February 20, 2012. The volunteers had no eye diseases as determined by examining their medical history and our examinations. The data from only the right eyes were analyzed. The inclusion criteria were: between 20- and 40-years old; eyes healthy according to slit–lamp biomicroscopy, ophthalmoscopy, and OCT; best-corrected visual acuity of less than or equal to 0.1 logMAR; and an IOP less than or equal to 21 mm Hg. The exclusion criteria were eyes with known ocular diseases such as glaucoma, staphyloma, and optic disc anomaly; eyes of subjects with known systemic diseases, such as hypertension and diabetes; presence of visual field defects; and prior refractive or intraocular surgery. None of the eyes was initially excluded because of poor OCT image quality caused by poor fixation. Seventy-two Japanese volunteers were screened for this project. One eye was excluded due to a superior segmental optic disc hypoplasia, and two eyes because of prior refractive surgery. Five other eyes were excluded because of segmentation error in the retinal thicknesses. In the end, the right eyes of 64 individuals (42 men and 22 women) were used for the analyses. 

Measurement of Axial Length, Refractive Error, and RT of 64 Sectors in Posterior Pole

All of the eyes had a standard ocular examination consisting of slit-lamp biomicroscopy of the anterior segment, ophthalmoscopy of the ocular fundus, IOP measurements with a pneumo-tonometer (CT-80; Topcon, Tokyo, Japan), and axial length measurements with the AL-2000 ultrasound instrument (TOMEY, Nagoya, Japan). The refractive error (spherical equivalent) was measured with the Topcon KR8800 auto-refractometer/keratometer. 

The retinal thickness was measured with the Spectralis SD-OCT (Heidelberg Engineering, Heidelberg, Germany) using the images obtained by PPAA scans. In this protocol, the OCT instrument automatically draws a line connecting the center of the fovea and the center of the optic disc as a reference line. Then, 61 line scans (1024 A scans/line) parallel to the reference line within the central 25° × 30° are recorded. In this protocol, B-scan observations were made by averaging five overlapping B-scans/image on each line scan. These five scans were used to obtain the RT values for each cell. The embedded OCT software determined the distance between the internal limiting membrane and the Bruch's membrane and recorded this distance as the RT. Then, three-dimensional RT maps were created based upon these data. The quality of the scans is indicated by a color scale at the bottom of the scanned images, and it had to be in the green range to be considered a good quality scan. In addition, the SD-OCT macular map provided a color scale representation of the topographic RT, which helped evaluate the image quality. Initially, we examined all of the B-scan images of each eye to determine whether there were any segmentation errors in the images. The following criteria were used to identify segmentation errors in the B-scans: obvious disruptions or abrupt 5% consecutive changes at a border or 20% cumulative of the entire image. ^7,12 Five eyes were excluded because of segmentation error in the RT. These excluded cases were perfectly matched by these two methods (i.e., the color scale and B-scan images). 

The average RT of each 8 × 8 (3° × 3°) sector, which made up the 64 sectors was determined. Our earlier study showed that the interobserver and the intervisit reproducibilities of the PPAA were excellent. ¹³ Therefore, one scan was sufficient to measure the mean RT of the 64 cells in the PPAA scan. The 64 cells were numbered as in Figure 1. 

Figure 1

 

The numbering of the 64 cells of the posterior pole asymmetry analysis.

Measurement of Position of Supra- and Infratemporall Peak RNFL Thickness Angles and Calculation of Difference Between Angles

The RNFL thickness was measured with the Spectralis SD-OCT using the RNFL circle scan. The temporal superior nasal inferior temporal (TSNIT) thickness curves were used to measure the angle between supra- or infratemporal peaks of the RNFL thickness and the fovea–disc axis. We determined the peak angles in the TSNIT thickness profile of the RNFL thickness analyses. The distance between the peak RNFL thickness for the supra- or infratemporal RNFL peaks and fovea–disc axis were determined by dragging a vertical line in the profile graph in the Spectralis OCT software and Photoshop (Adobe, San Jose, CA, USA). The position of fovea–disc axis was determined as the intersection between the fovea–disc axis of the PPAA scan and the scan circle of RNFLT. Then, the distance between the fovea–disc axis and supra- (X1a + X1b of Fig. 2) or infratemporal RNFL peaks (X2 of Fig. 2) was converted to an angular value by dividing by the entire distance (Y) and multiplying by 360 (Fig. 2). ^14,15 The peak angle difference (PAD) was calculated as the difference between the supratemporal peak RNFL angle and the infratemporal peak RNFL angle. 

Figure 2

 

Measurement of the angle between the supra- and infrapeak RNFL thickness.

Statistical Analyses

All statistical analyses were performed with the SPSS statistics 21 for Windows (SPSS, Inc., IBM, Somers, New York, USA). The relationships between the 32 corresponding upper and lower RT pairs were determined by the Pearson coefficients of correlation. The relationships between the RT of each of the difference between corresponding upper and lower RT of 32 pairs and the PAD were determined by the Spearman coefficients of correlation. 

Results

The mean ± SD of the age was 26.0 ± 4.5 years (range, 22–40 years), and the mean refractive error (spherical equivalent) was −4.1 ± 3.1 diopters (D) (range, −14.3 to 0.0 D). The mean axial length was 25.0 ± 1.3 mm (range, 22.4–28.2 mm), and the mean PAD was −1.9° ± 11.5° (range, −27.4° to 23.5°). The differences between corresponding upper and lower RT of the 32 pairs are shown in Figure 3. 

Figure 3

 

Differences of retinal thickness between corresponding upper and lower 32 pairs of sectors.

Pearson Coefficients of Correlation Between Corresponding Upper and Lower RT Cells

The RTs of the corresponding upper and lower 32 pairs of cells were significantly correlated (R = 0.45–0.97, P < 0.001; Table). The correlation coefficients of the pairs of central (12° × 12°) and temporal macular cells were higher than that of the pairs of nasal-macular cells. 

Table.

 

Difference and Correlation Between Corresponding Upper and Lower Cells Retinal Thickness and Its Relationship With Peak Angle Difference

Corresponding Cell Number	Difference Between Corresponding Upper and Lower Cells Retinal Thickness, mm			Pearson Correlation Coefficient		Spearman Correlation Coefficient
				Corresponding Upper and Lower Cells Retinal Thickness		Difference Between Corresponding Cells Retinal Thickness and Peak Angle Difference
	Mean ± SD	Range	Absolute Value Mean ± SD	R	P Value	R	P Value
1	3.2 ± 8.3	−19 ∼ 25	6.6 ± 5.7	0.74	<0.001*	0.04	0.79
2	4.1 ± 9.2	−11 ∼ 28	8.1 ± 6.0	0.75	<0.001*	0.07	0.60
3	3.7 ± 10.3	−21 ∼ 27	8.7 ± 6.5	0.72	<0.001*	0.05	0.72
4	2.7 ± 12.2	−26 ∼ 27	10.2 ± 7.1	0.66	<0.001*	0.18	0.15
5	2.8 ± 12.8	−26 ∼ 27	10.5 ± 7.8	0.70	<0.001*	0.16	0.22
6	−3.9 ± 14.2	−39 ∼ 24	11.8 ± 8.6	0.68	<0.001*	0.14	0.28
7	2.8 ± 15.1	−34 ∼ 43	11.9 ± 9.7	0.64	<0.001*	0.11	0.39
8	21.5 ± 21.0	−21 ∼ 66	23.2 ± 18.8	0.45	<0.001*	0.06	0.66
9	5.4 ± 7.8	−13 ∼ 23	7.6 ± 5.6	0.77	<0.001*	−0.14	0.27
10	8.0 ± 9.1	−18 ∼ 25	10.1 ± 6.7	0.74	<0.001*	−0.08	0.53
11	9.6 ± 8.3	−7 ∼ 31	10.1 ± 7.7	0.80	<0.001*	0.01	0.93
12	12.2 ± 7.6	−4 ∼ 31	12.4 ± 7.2	0.85	<0.001*	0.22	0.08
13	14.8 ± 8.4	−6 ∼ 39	15.0 ± 8.0	0.84	<0.001*	0.25	0.04*
14	7.9 ± 10.1	−24 ∼ 30	10.7 ± 7.1	0.80	<0.001*	0.43	<0.001*
15	−5.0 ± 14.3	−54 ∼ 18	11.4 ± 9.9	0.72	<0.001*	0.45	<0.001*
16	−2.5 ± 16.1	−34 ∼ 29	13.3 ± 9.1	0.68	<0.001*	0.35	0.01*
17	4.9 ± 7.4	−13 ∼ 27	6.8 ± 5.7	0.81	<0.001*	0.07	0.56
18	5.9 ± 7.1	−12 ∼ 26	7.2 ± 5.8	0.86	<0.001*	0.21	0.10
19	3.8 ± 8.0	−22 ∼ 31	6.5 ± 5.9	0.86	<0.001*	0.08	0.52
20	5.7 ± 7.3	−15 ∼ 25	7.3 ± 5.6	0.89	<0.001*	0.10	0.43
21	8.9 ± 8.4	−6 ∼ 27	9.6 ± 7.6	0.87	<0.001*	0.19	0.14
22	7.1 ± 8.4	−13 ∼ 26	8.8 ± 6.6	0.87	<0.001*	0.25	0.05*
23	6.7 ± 9.0	−11 ∼ 26	8.9 ± 6.8	0.84	<0.001*	0.32	0.01*
24	−3.1 ± 15.9	−43 ∼ 32	12.8 ± 9.4	0.74	<0.001*	0.40	<0.01*
25	0.2 ± 4.3	−13 ∼ 10	3.3 ± 2.7	0.95	<0.001*	−0.05	0.71
26	−0.6 ± 4.2	−11 ∼ 9	3.3 ± 2.7	0.96	<0.001*	−0.01	0.96
27	−3.9 ± 3.9	−16 ∼ 3	4.2 ± 3.6	0.97	<0.001*	0.04	0.76
28	−5.5 ± 11.3	−33 ∼ 15	10.0 ± 7.6	0.77	<0.001*	0.24	0.06
29	−0.5 ± 9.1	−26 ∼ 17	7.1 ± 5.7	0.86	<0.001*	0.11	0.41
30	0.8 ± 4.0	−13 ∼ 9	3.1 ± 2.7	0.97	<0.001*	0.16	0.22
31	4.8 ± 5.9	−14 ∼ 28	5.8 ± 4.8	0.94	<0.001*	0.27	0.03*
32	6.3 ± 11.1	−25 ∼ 47	8.6 ± 8.9	0.81	<0.001*	0.05	0.69

*  Statistically significant correlation (P < 0.05).

Spearman Correlation Coefficient of Differences Between RT of Corresponding Cells and PAD

The PAD was significantly correlated with the differences between corresponding RTs of the eight nasal pairs (R = 0.25–0.45, P < 0.05) but was not significantly correlated with the differences between the corresponding RTs of the central and the 24 temporal macular pairs (P > 0.05; Table). 

Discussion

Our results showed that the RT of cells in the temporal and central areas of the central fundus were quite similar to the corresponding cells across the fovea–disc axis, indicating that the symmetry was good in these areas. The differences in the thicknesses between the upper and lower counterparts across the fovea–disc axis was mostly less than 30 μm (2000/2048 cells, 97.7%). Seo et al. ⁷ reported that the diagnostic ability of PPAA for detecting localized RNFL defects was good with good sensitivity. In their algorithm, a difference of 30 μm or greater was used as the cutoff value for the cell-to-cell comparisons. The area under the receiver operating characteristic of the PPAA based on the number of black cells was 0.958 ± 0.013 in 84 open-angle glaucoma subjects with localized, wedge-shaped-RNFL defects in the red-free RNFL photographs and 122 eyes of healthy subjects. ⁷ This cutoff value was selected based on the findings reported by previous studies. ^8,9 Because the symmetry of upper and lower cells was good with the values generally less than 30 μm in healthy eyes in our study, this value would be reasonable and acceptable. Additionally, because the differences of the corresponding cells in the lateral macular area (cell numbers 25, 26, and 27) was very small in healthy eyes, even a 10-μm cutoff value might be acceptable in identifying an alteration of the symmetry for cells of that area in early glaucomatous eyes. For example, Um et al. ¹⁰ divided the macular thickness values into those corresponding to the superior and inferior macular thickness zones, and then evaluated the symmetry between corresponding superior and inferior macular zones. They considered that the results of these comparisons were more sensitive than the RNFL thickness for detecting early-stage glaucoma. The cutoff values of the zones, as obtained by comparison with reference group data varied, and the cutoff values of the temporal zone (cell number 25, 26, and 27) were the smallest (7.7 μm). ¹⁰  

However, the symmetry was not well preserved in the nasal-peripheral macular areas (number of corresponding cells: 1–10, 15, 16, 24), and the correlation coefficient of nasal-peripheral macular area was less than that of the central- and temporal-macular areas (number of corresponding cells: 11–14, 17–23, 25–32; Fig. 4). In some cases, the differences were found to be greater than 30 μm, and these eyes had asymmetry of the thick retinal nerve fiber bundles in the superior and inferior sectors (Fig. 5). 

Figure 4

 

Correlation coefficients of each pair are shown in the upper cells. The gray tone also indicates the value of each cell as on the right scale bar. The correlation coefficient equal to or lower than 0.7 are shown in white. T, temporal; F, fovea; N, nasal.

Figure 5

 

(A, B) Grid of 64 cells superimposed on fundus photographs. The thickness of each cell is designated numerically and also color-coded. The fovea–disc axis is shown by the slanted green-blue line. The supratemporal and infratemporal peak RNFL positions are shown by the yellow dots. Supra- and infrapeak RNFL angles are shown by yellow two-way arrows. (C, D) Hemisphere asymmetry analysis obtained by posterior pole asymmetry analysis. (E, F) The supra- and infratemporal peak RNFL positions are shown by the black lines and the fovea–disc axis is shown by the blue line in the RNFL profile. The sites of the angles of the supra- and infratemporal peak thicknesses of the RNFL against the fovea–disc axis were approximately the same for case shown in (A, C, E) and different for case shown in (B, D, F).

We assumed that the structural asymmetry between the supra- and infratemporal RNFL bundles across the fovea–disc axis may be the cause of these differences. Thus, we compared the difference of the PAD across the fovea–disc axis and the differences of the corresponding cells. Our findings showed that the differences of the corresponding cells (e.g., paired-cells such numbers as 13–16, and 22–24, and 31) were significantly and positively correlated with the difference of the PAD (Fig. 5). This is a new finding and can affect the accuracy of the diagnosis of early-stage glaucomatous eyes. 

An earlier study showed that the sectorial retinal thickness was significantly correlated with the axial length. ¹³ Thus, the axial length may affect the relationship between the superior and inferior retinal thickness. A partial correlation analysis was performed to determine the relationship between the superior and inferior retinal thicknesses with an exclusion of the effect of the axial length. The results showed that even after excluding the effect of the axial length, the corresponding superior and inferior cells retinal thickness were highly correlated (R = 0.30–0.96, P = 0.02 to <0.001, Supplementary Table S1). 

In the earlier study of the symmetry of retinal thicknesses between the upper- and lower-lateral areas, the correlation coefficient was 0.335 or a coefficient of determination of 0.112, which is smaller than our results with a correlation coefficient of 0.45 to 0.97. It is possible that the area studied earlier was larger than the area studied in this study. This is important because the larger the area studied, the greater will be the variations of the retinal thicknesses. Additionally, our subjects were young and had healthy eyes. In addition, the ageing changes and media opacities were minimal, which could account for the higher coefficients of correlation. 

Because the location of the large superior and inferior vessels are supposedly located at the sites of the peak RNFL, ^15–17 and because these vessels are not necessarily symmetrical across the fovea–disc axis, their locations can affect the symmetry of upper and lower retinal thicknesses. We measured the supra- and infratemporal artery and vein angles in the RNFL thickness scan circle using the same method as the peak RNFL locations. The artery angle difference (AAD) and the vein angle difference (VAD) were calculated as the difference between the supratemporal artery or vein angles and the infratemporal artery or vein angles. The peak angle difference (PAD) was significantly correlated with the AAD (R = 0.44, P < 0.001) and the VAD (R = 0.32, P = 0.012). The AAD was significantly correlated with the differences between corresponding RTs of the five nasal pairs (R = 0.26–0.33, P < 0.05) but was not significantly correlated with the differences between the corresponding RTs of the other pairs (P > 0.05; Supplementary Table S2). These correlations on the relationship between the PAD and the difference between corresponding RTs are weak. The VAD was significantly correlated with the differences between corresponding RTs of only one pair (sector 31, R = 0.27, P = 0.03; Supplementary Table S2). These results suggest that the AAD can be substituted for the PAD in assessing the symmetry between upper and lower retinal thicknesses. They also indicate that the asymmetry of the superior and inferior retinal artery position of the peripapillary area across the fovea–disc axis would break the symmetry of the RT across the fovea–disc axis. 

The angle of the fovea–disc axis may be correlated with the PAD. Therefore, we investigated the fovea–disc angle and the relationship between the PAD and the differences between corresponding RTs. The fovea–disc angle was calculated by dividing the fovea disc axis (X1b of Fig. 2) by the entire distance and multiplying by 360 (Fig. 2). The fovea–disc angle was significantly correlated with the differences between corresponding RTs of the eight nasal pairs (R = 0.26–0.42, P < 0.05) and one peripheral pair (R = 0.28, P = 0.02), but was not significantly correlated with the differences between the corresponding RTs of the other pairs (P > 0.05; Supplementary Table S3). These results are similar to the correlations between the PAD and the differences of the corresponding RTs. In addition, the fovea–disc angle was significantly correlated with the PAD (R = −0.43, P < 0.001). These results suggest that the fovea–disc angle may be substituted for the PAD in assessing the symmetry between upper and lower retinal thicknesses in the different sectors. 

The earliest sign of glaucoma is usually a focal reduction of the retinal thickness of the RNFL bundle especially in the superior or inferior lateral areas. The structural asymmetry of these areas might interfere with the diagnostic accuracy of PPAA. Indeed, Seo et al. ⁷ reported that the sensitivity and specificity of PPAA in detecting localized RNFL defects using two, three, and four consecutive black cells (black indicating differences of retinal thickness of >30 μm) were 95.2% vs. 81.1%, 83.3% vs. 92.6%, and 69.0% vs. 98.4%. Additionally, Sullivan-Mee et al. ¹⁸ reported that the sensitivity and specificity of the PPAA for detecting early-stage glaucoma using absolute differences between the overall superior and inferior macular thicknesses, were 77.3% and 80.0%, respectively. Thus, a simple symmetry theory does not necessarily satisfy both of sensitivity and specificity of the PPAA analysis. However, modifying the asymmetry analyses using the PAD should be considered. 

This study has several limitations. First, the study was not a population-based study. Epidemiologic studies have shown that the Japanese population is one of the most myopic groups, ¹⁹ and the present study group was university students who are known to be myopic. Thus, our results describe the characteristics of young myopic eyes, but might not necessarily hold for older and nonmyopic populations. On the other hand, the reliability of the examination is very high because no pathological factors such as cataract or vitreal opacities were present in the young healthy individuals and the understanding of the examination procedures was high. Furthermore, the narrow range of age prevented the study from interference of the cohort effects and the age effects. A second limitation was that segmentation errors were found in cell number 32 (lateral of optic disc) in 20 eyes out of 64 eyes. Most of these eyes had an optic disc conus. Therefore, the reliability of the measurements of retinal thickness were not good for this cell. However, there were few segmentation errors for other cells. The third limitation of this study was that the circular scan was not adjusted for ocular magnification. According to studies on scan circle size and RNFL thickness, ^20,21 this may have introduced some errors in the mean RNFL thickness profiles and the mean angles of maximums determined. However, considering the extension of the peripapillary retinal nerve fibers, the angles of peak thicknesses seem to be less affected by ocular magnification. To the best of our knowledge, this is the first study to investigate the symmetry between the RT of corresponding cells by PPAA. However, additional studies with a larger sample size and broad range of ages are needed to determine whether there are patterns in the symmetry between individuals corresponding to the upper and lower RT of the 64 cells. 

In summary, we found by PPAA that the symmetry of the RT between corresponding upper and lower cells was high in the central and temporal macular areas, but not so high in the peripheral and nasal-macular area. The structural asymmetry of superior and inferior RNFL bundles of the nasal macular area would break the symmetry of the RT across the macula-disc axis. This information would be important for interpreting the PPAA for diagnosing early glaucoma. 

Supplementary Materials

pdf S1 

Acknowledgments

Disclosure: T. Yamashita, None; T. Sakamoto, None; N. Kakiuchi, None; M. Tanaka, None; Y. Kii, None; K. Nakao, None