Usefulness of Automated Measurements of Localized Retinal Nerve Fiber Layer Defects Area Using Significance Map

Sam Seo; Joong Won Shin; Ki Bang Uhm

doi:10.3341/jkos.2013.54.6.902

Journal List > J Korean Ophthalmol Soc > v.54(6) > 1009702

Go to TopGo to Top Go to BottomGo to Bottom

TOOLS

Seo, Shin, and Uhm: Usefulness of Automated Measurements of Localized Retinal Nerve Fiber Layer Defects Area Using Significance Map

Original Article

J Korean Ophthalmol Soc 2013;54(6):902-912.

Published online: 7 September 2013

DOI: https://doi.org/10.3341/jkos.2013.54.6.902

Usefulness of Automated Measurements of Localized Retinal Nerve Fiber Layer Defects Area Using Significance Map

Sam Seo, MD, Joong Won Shin, MD, Ki Bang Uhm, MD, PhD

Department of Ophthalmology, Hanyang University College of Medicine, Seoul, Korea

Address reprint requests to Ki Bang Uhm, MD, PhD Department of Ophthalmology, Hanyang University Seoul Hospital #222 Wangsimni-ro, Seongdong-gu, Seoul 133-791, Korea Tel: 82-2-2290-8570, Fax: 82-2-2291-8517, E-mail: KBUhm@hanyang.ac.kr

Received 30 November 201228 December 2012 Accepted 8 April 2013

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Purpose

To evaluate the usefulness of automated measurements of the localized retinal nerve fiber layer (RNFL) defects area in patients with glaucoma.

Methods

Fifty one patients with localized RNFL defects in RNFL red-free photographs and 53 healthy subjects were included. All participants were imaged with 3D spectral-domain optical coherence tomography (OCT). The area of defects was measured with the RNFL significance map (red = p < 1% and yellow = p < 5%) using Image J manually and Matlab software automatically. The area under the receiver operating characteristic curve (AUC) was calculated for the RNFL de-fect area of the RNFL photograph and RNFL maps, circumpapillary RNFL thickness, optic disc parameter, and macular in-ner retina thickness.

Results

High correlation was observed between manually and automatically measured defect areas in the significance map (red area r = 0.904, red and yellow area r = 0.890). The AUC for manually and automatically measured defects area (0.987, 0.966; p < 5%, p = 0.31, respectively) in the significance map was comparable. The latter demonstrated slightly higher but insignificant difference in AUC for inferior quadrant circumpapillary RNFL thickness (0.936; p = 0.22) and was significantly higher than the inferior ganglion cell layer plus inner plexiform layer thickness (0.894) and vertical cup to disc ratio (0.869) (p = 0.018, p = 0.008, respectively).

Conclusions

The automated measurements of the RNFL defect area in the significance map performed adequately in de-tecting localized RNFL defects and had a better performance than macular inner retina and optic nerve parameters.

Keywords: Glaucoma, Optic nerve, Retinal nerve fiber layer, Significance map, Spectral-domain optical coherence tomography

References

1. Schuman JS. Spectral domain optical coherence tomography for glaucoma (an AOS thesis). Trans Am Ophthalmol Soc. 2008; 106:426–58.

2. Kim NR, Lee ES, Seong GJ. . Spectral-domain optical coher-ence tomography for detection of localized retinal nerve fiber layer defects in patients with open-angle glaucoma. Arch Ophthalmol. 2010; 128:1121–8.

3. Leung CK, Lam S, Weinreb RN. . Retinal nerve fiber layer imaging with spectral-domain optical coherence tomography: analysis of the retinal nerve fiber layer map for glaucoma detection. Ophthalmology. 2010; 117:1684–91.

4. Leung CK, Choi N, Weinreb RN. . Retinal nerve fiber layer imaging with spectral-domain optical coherence tomography: pat-tern of RNFL defects in glaucoma. Ophthalmology. 2010; 117:2337–44.

5. Kang SY, Sung KR, Na JH. . Comparison between deviation map algorithm and peripapillary retinal nerve fiber layer measure-ments using Cirrus HD-OCT in the detection of localized glau-comatous visual field defects. J Glaucoma. 2012; 21:372–8.

6. Jeoung JW, Park KH, Kim TW. . Diagnostic ability of optical coherence tomography with a normative database to detect lo-calized retinal nerve fiber layer defects. Ophthalmology. 2005; 112:2157–63.

7. Yoo YC, Park KH. Comparison of optical coherence tomography and scanning laser polarimetry for detection of localized retinal nerve fiber layer defects. J Glaucoma. 2010; 19:229–36.

8. Yoo YC, Park KH. Influence of angular width and peripapillary position of localized retinal nerve fiber layer defects on their de-tection by time-domain optical coherence tomography. Jpn J Ophthalmol. 2011; 55:115–22.

9. Kang SM, Uhm KB. False negative findings of optical coherence tomography in eyes with localized nerve fiber layer defects. J Korean Ophthalmol Soc. 2011; 52:454–61.

10. Ye C, To E, Weinreb RN. . Comparison of retinal nerve fiber layer imaging by spectral domain optical coherence tomography and scanning laser ophthalmoscopy. Ophthalmology. 2011; 118:2196–202.

11. Mwanza JC, Oakley JD, Budenz DL, Anderson DR. Cirrus Optical Coherence Tomography Normative Database Study Group. Ability of cirrus HD-OCT optic nerve head parameters to discriminate nor-mal from glaucomatous eyes. Ophthalmology. 2011; 118:241–8. e1.

12. Schulze A, Lamparter J, Pfeiffer N. . Diagnostic ability of reti-nal ganglion cell complex, retinal nerve fiber layer, and optic nerve head measurements by Fourier-domain optical coherence tomography. Graefes Arch Clin Exp Ophthalmol. 2011; 249:1039–45.

13. Seong M, Sung KR, Choi EH. . Macular and peripapillary reti-nal nerve fiber layer measurements by spectral domain optical co-herence tomography in normal-tension glaucoma. Invest Ophthalmol Vis Sci. 2010; 51:1446–52.

14. Tan O, Li G, Lu AT. . Mapping of macular substructures with opti-cal coherence tomography for glaucoma diagnosis. Ophthalmology. 2008; 115:949–56.

15. Taliantzis S, Papaconstantinou D, Koutsandrea C. . Comparative studies of RNFL thickness measured by OCT with global index of visual fields in patients with ocular hypertension and early open angle glaucoma. Clin Ophthalmol. 2009; 3:373–9.

16. Cvenkel B, Kontestabile AS. Correlation between nerve fibre layer thickness measured with spectral domain OCT and visual field in patients with different stages of glaucoma. Graefes Arch Clin Exp Ophthalmol. 2011; 249:575–84.

17. Lopez-Peña MJ, Ferreras A, Larrosa JM. . Relationship be-tween standard automated perimetry and retinal nerve fiber layer parameters obtained with optical coherence tomography. J Glaucoma. 2011; 20:422–32.

18. Harwerth RS, Charles F. Prentice Award Lecture 2006: a neuron doctrine for glaucoma. Optom Vis Sci. 2008; 85:436–44.

19. Tan O, Chopra V, Lu AT. . Detection of macular ganglion cell loss in glaucoma by Fourier-domain optical coherence tomography. Ophthalmology. 2009; 116:2305–14. e1-2.

20. Mori S, Hangai M, Sakamoto A, Yoshimura N. Spectral-domain optical coherence tomography measurement of macular volume for diagnosing glaucoma. J Glaucoma. 2010; 19:528–34.

21. Nakatani Y, Higashide T, Ohkubo S. . Evaluation of macular thickness and peripapillary retinal nerve fiber layer thickness for detection of early glaucoma using spectral domain optical coher-ence tomography. J Glaucoma. 2011; 20:252–9.

22. Mwanza JC, Durbin MK, Budenz DL. . Glaucoma diagnostic accuracy of ganglion cell-inner plexiform layer thickness: compar-ison with nerve fiber layer and optic nerve head. Ophthalmology. 2012; 119:1151–8.

Figure 1.

(A) The cSLO RNFL photographs were overlaid with a color disc photograph (5 × 5 mm²) as a reference image that was obtained by Topcon 3D OCT-2000. (B) The retinal blood vessels (arrow) and RNFL defects (arrow head) were matched between the cSLO RNFL photograph and the reference image. cSLO = confocal scanning laser ophthalmoscope; RNFL = retinal nerve fi-ber layer; OCT = optical coherence tomography; T = temporal; N = nasal.

Figure 2.

(A) Original photograph in the retinal nerve fiber layer (RNFL) significance map, RNFL thickness map, optic disc photo-graph, and cSLO RNFL photograph. The RNFL significance map, RNFL thickness map, and optic disc photograph were obtained by Topcon 3D OCT-2000, and the RNFL photograph was captured by a Nidek F-10 cSLO. (B) Each boundary of the RNFL defects was manually delineated. The areas of the RNFL defects were measured on the 5 × 5 mm² parapapillary area using Image J software (yellow lines) manually. Superpixels coded in red, red and yellow were automatically calculated by a Matlab computer program ac-cording to the red, green, and blue values in the 5 × 5 mm² circumpapillary region. Measurement of the area of localized RNFL defect in the significance map (red area = 1.11 mm², red and yellow area = 2.79 mm²), thickness map (1.40 mm²), cSLO RNFL photograph (1.77 mm²) manually, and automated measurement using significance map (red area = 0.84 mm², red and yellow area = 3.59 mm²). cSLO = confocal scanning laser ophthalmoscope; RNFL = retinal nerve fiber layer; OCT = optical coherence tomography.

Figure 3.

(A) Scatter plots showing the relationship of manually measured red area (p < 1%) of significance map against automati-cally measured red area (p < 1%) of significance map (r = 0.904, p < 0.001). (B) Relationship of manually measured red and yel-low area (p < 5%) of significance map against automatically measured red and yellow area (p < 5%) of significance map (r = 0.890, p < 0.001).

Figure 4.

Bland-Altman plots of agreement between manually and automatically measured (A) red area (p < 1%) and (B) red and yellow area (p < 5%) in the significance map. Differences between automatically and manually measured areas were plotted against the means. Dash lines represent the 95% confidence intervals.

Figure 5.

The cut-off point (>0.546 mm²), sensitivity (90.2%), and specificity (94.3%) of automated measurements of sig-nificance map red and yellow area (p < 5%).

Figure 6.

Receiver operating characteristics curves of the man-ually measured significance map red and yellow area (p < 5%), automatically measured significance map red and yellow area (p < 5%), circumpapillary inferior quadrant retinal nerve fi-ber layer (RNFL) thickness, inferior ganglion cell layer + in-ner plexiform layer thickness (GCL + IPL thickness), and vertical cup to disc ratio (CDR).

Table 1.

Demographics and characteristics of subjects with localized RNFL defects and normal controls

	Normal eyes (n = 53)	Eyes with RNFL defects (n = 51)	p-value
Age (years)	56.3 ± 10.7	57.7 ± 13.7	0.54^*
Gender (male:female)	28:25	29:22	0.68^†
Eye (right:left)	24:29	21:30	0.67^†
Spherical equivalent (diopters)	-0.15 ± 1.29	-0.65 ± 2.05	0.15^*
Central corneal thickness (μ m)	543.78 ± 32.17	539.33 ± 31.98	0.50^*
Mean deviation (dB)	-1.45 ± 0.91	-6.15 ± 4.33	<0.001^*
Pattern standard deviation (dB)	1.58 ± 0.69	6.08 ± 4.20	<0.001^*
Type of glaucoma	NA
Preperimetric		7
Perimetric		44
Localized RNFL defects	NA	64
Superior		11
Inferior		27
Superior and inferior		13

Values are presented as number or mean ± SD.

RNFL = retinal nerve fiber layer; NA = not applicable.

^* Unpaired t-test;

^† Chi-square test.

Table 2.

Area under the receiver operating characteristic curves (AUCs) of the red-free photograph, thickness map, and significance map between normal controls and eyes with RNFL defects

	Normal (n = 53)	Eyes with RNFL defects (n = 51)	p-value^*	AUCs (95% CI)	Sn/Sp (Sp ≥ 80%)	Sn/Sp (Sp ≥ 90%)
Defect area (mm²)
Photographic (manual)	0	2.41 ± 1.46	<0.001	1.000 (1.000-1.000)	NA	100/100
Thickness map (manual)	0	2.04 ± 1.35	<0.001	1.000 (1.000-1.000)	NA	100/100
Significance map red (manual)	0	1.65 ± 1.17	<0.001	0.961 (0.924-0.998)	NA	92/100
Significance map red & yellow (manual)	0.01 ± 0.10	2.77 ± 1.83	<0.001	0.987 (0.967-1.000)	NA	98/98
Significance map red (automatic)	0.01 ± 0.04	1.70 ± 1.43	<0.001	0.963 (0.925-1.000)	94/85	94/92
Significance map red & yellow (automatic)	0.16 ± 0.32	3.15 ± 2.25	<0.001	0.966 (0.932-1.000)	98/83	90/94

RNFL = retinal nerve fiber layer; CI = confidence interval; Sn = sensitivity; Sp = specificity; NA = not applicable.

^* Unpaired t-test.

Table 3.

Area under the receiver operating characteristic curves (AUCs) of circumpapillary RNFL thickness between normal con-trols and eyes with RNFL defects

	Normal (n = 53)	Eyes with RNFL defects (n = 51)	p-value^*	AUCs (95% CI)	Sn/Sp (Sp ≥ 80%)	Sn/Sp (Sp ≥ 90%)
Circumpapillary RNF	FL thickness (μ m)
Average	104.2 ± 8.1	83.3 ± 12.8	<0.001	0.917 (0.864-0.970)	78/81	76/92
Quadrant
Temporal	79.2 ± 10.6	69.3 ± 12.2	<0.001	0.726 (0.628-0.823)	55/81	35/91
Superior	122.7 ± 12.6	101.7 ± 18.3	<0.001	0.863 (0.791-0.935)	80/85	67/91
Nasal	84.7 ± 14.4	74.5 ± 15.8	0.001	0.710 (0.608-0.811)	51/81	39/91
Inferior	129.2 ± 10.8	90.3 ± 22.2	<0.001	0.936 (0.890-0.981)	88/81	82/91
Clock-hour
9 (temporal)	68.0 ± 10.9	64.3 ± 12.6	0.12	0.602 (0.492-0.713)	35/81	24/91
10	91.2 ± 12.8	77.1 ± 19.4	<0.001	0.719 (0.618-0.820)	57/81	47/91
11	127.2 ± 17.5	99.6 ± 29.2	<0.001	0.786 (0.695-0.877)	65/81	61/91
12 (superior)	122.6 ± 18.2	103.3 ± 22.9	<0.001	0.771 (0.680-0.862)	51/81	39/92
1	122.0 ± 15.2	102.3 ± 19.3	<0.001	0.795 (0.709-0.881)	65/81	49/91
2	99.1 ± 16.2	85.3 ± 20.5	<0.001	0.718 (0.617-0.819)	61/81	33/92
3 (nasal)	74.4 ± 15.5	66.4 ± 15.3	0.009	0.667 (0.562-0.773)	41/81	20/92
4	80.3 ± 18.2	71.6 ± 19.6	0.02	0.644 (0.537-0.751)	41/81	20/91
5	114.1 ± 16.5	95.3 ± 20.5	<0.001	0.767 (0.674-0.861)	67/83	41/91
6 (inferior)	134.4 ± 17.5	93.4 ± 25.7	<0.001	0.896 (0.836-0.957)	80/83	75/91
7	138.7 ± 17.5	82.0 ± 34.6	<0.001	0.907 (0.845-0.970)	84/85	82/94
8	78.3 ± 12.9	66.6 ± 16.2	<0.001	0.736 (0.637-0.835)	63/81	49/92

RNFL = retinal nerve fiber layer; CI = confidence interval; Sn = sensitivity; Sp = specificity.

^* Unpaired t-test.

Table 4.

Area under the receiver operating characteristic curves (AUCs) of macular inner retina thickness and optic disc parameter between normal controls and eyes with RNFL defects

	Normal (n = 53)	Eyes with RNFL defects (n = 51)	p-value^*	AUCs (95% CI)	Sn/Sp (Sp ≥ 80%)	Sn/Sp (Sp ≥ 90%)
Macular inner retina thick	kness (μ m)
NFL
Superior	34.1 ± 4.9	29.6 ± 7.5	<0.001	0.719 (0.616-0.822)	59/81	39/92
Inferior	37.6 ± 4.2	24.3 ± 10.5	<0.001	0.859 (0.778-0.941)	75/83	73/92
Total	35.8 ± 4.0	26.9 ± 7.6	<0.001	0.846 (0.762-0.930)	78/83	73/94
GCL + IPL
Superior	72.1 ± 5.7	65.9 ± 5.5	<0.001	0.789 (0.702-0.876)	51/85	39/94
Inferior	70.5 ± 5.9	60.4 ± 6.2	<0.001	0.894 (0.832-0.956)	75/85	71/91
Total	71.3 ± 4.8	63.2 ± 5.0	<0.001	0.879 (0.815-0.943)	65/83	59/94
GCC
Superior	106.2 ± 7.8	95.5 ± 12.0	<0.001	0.770 (0.678-0.863)	61/83	39/91
Inferior	108.0 ± 8.3	84.6 ± 15.6	<0.001	0.893 (0.829-0.958)	82/81	73/91
Total	107.2 ± 7.8	90.1 ± 11.9	<0.001	0.875 (0.806-0.943)	82/81	67/92
Optic nerve head
Cup area (mm²)	1.06 ± 0.50	1.61 ± 0.68	<0.001	0.732 (0.636-0.828)	47/83	43/91
Rim area (mm²)	2.01 ± 0.47	1.24 ± 0.52	<0.001	0.869 (0.796-0.941)	84/85	65/91
C/D area ratio	0.34 ± 0.14	0.56 ± 0.17	<0.001	0.827 (0.747-0.907)	69/81	63/91
Linear CDR	0.57 ± 0.14	0.74 ± 0.13	<0.001	0.824 (0.744-0.905)	69/81	61/91
Vertical CDR	0.58 ± 0.13	0.79 ± 0.14	<0.001	0.869 (0.798-0.940)	NA	73/91
Cup volume (mm³)	0.24 ± 0.18	0.35 ± 0.25	0.01	0.628 (0.520-0.736)	43/81	18/91
Rim volume (mm³)	0.17 ± 0.19	0.15 ± 0.17	0.56	0.562 (0.448-0.675)	37/83	22/91

RNFL = retinal nerve fiber layer; CI = confidence interval; Sn = sensitivity; Sp = specificity; NFL = nerve fiber layer; GCL = ganglion cell layer; IPL = inner plexiform layer; GCC = ganglion cell complex; CDR = cup to disc ratio; NA = not applicable.

^* Unpaired t-test.

Table 5.

Comparison of significance map area analyzed by manual and automatic method

	Manual	Automatic	Difference	p-value^*	r	p-value^†
Significance map red area (mm²)	0.81 ± 1.16	0.84 ± 1.31	-0.03 ± 0.56	0.54	0.904	<0.001
Significance map red and yellow area (mm²)	1.38 ± 1.89	1.64 ± 2.19	-0.26 ± 1.00	0.009	0.890	<0.001

^* Paired t-test;

^† Correlation

Table 6.

Relationships (correlation coefficients) between retinal nerve fiber layer defect areas and structural-functional p parameters

	RNFL defects area
	Red-free Photograph (manual)	Thickness Map (manual)	p < 1%(manual)	p < 1 & 5% (manual)	p < 1%(automatic)	p < 1 & 5% (automatic)	MD	PSD
RNFL defects area
Photograph (manual)		0.962	0.905	0.902	0.798	0.817	-0.821	0.816
Thickness map (manual)	0.962		0.929	0.904	0.818	0.831	-0.822	0.836
p < 1% (manual)	0.905	0.929		0.957	0.904	0.905	-0.808	0.819
p < 1 & 5% (manual)	0.902	0.904	0.957		0.842	0.890	-0.788	0.792
p < 1% (automatic)	0.798	0.818	0.904	0.842		0.970	-0.717	0.708
p < 1 & 5% (automatic)	0.817	0.831	0.905	0.890	0.970		-0.723	0.721
VF indices
MD	-0.821	-0.822	-0.808	-0.788	-0.717	-0.723		-0.862
PSD	0.816	0.836	0.819	0.792	0.708	0.721	-0.862

All correlations are significant at the p < 0.001 level (two-tailed).

RNFL = retinal nerve fiber layer; VF = visual field; MD = mean deviation; PSD = pattern standard deviation.

Table 7.

Probability values for the differences (upper triangle of numbers) between area under the receiver operating characteristic curves

(AUC)	Significance map red manual (0.961)	Significance map red & yellow manual (0.987)	Significance map red auto (0.963)	Significance map red & yellow auto (0.966)	RNFLT e inferior quadrant (0.936)	Macular inner retina thickness Inferior GCL + IPL (0.894)	Vertical CDR (0.869)
Significance map red manual (0.961)	-	0.21	0.92	0.79	0.37	0.04	0.004
Significance map red & yellow manual (0.987)		-	0.26	0.31	0.04	0.005	0.002
Significance map red auto (0.963)			-	0.77	0.34	0.04	0.007
Significance map red & yellow auto (0.966)				-	0.22	0.02	0.008
RNFLT inferior quadrant (0.936) Macular inner retina thickness					-	0.15	0.07
Inferior GCL+IPL (0.894)						-	0.45
Vertical CDR (0.869)							-

AUC = area under the receiver operating characteristic curve; RNFLT = retinal nerve fiber layer thickness; GCL = ganglion cell layer; IPL = inner plexiform layer; CDR = cup to disc ratio.

TOOLS

Similar articles