Limits...
Three-dimensional spectral-domain optical coherence tomography data analysis for glaucoma detection.

Xu J, Ishikawa H, Wollstein G, Bilonick RA, Folio LS, Nadler Z, Kagemann L, Schuman JS - PLoS ONE (2013)

Bottom Line: Features of super pixel map were extracted and used as inputs to machine classifier (LogitBoost adaptive boosting) to automatically identify diseased eyes.The super pixel analysis showed statistically significantly higher AUC than the cpRNFL (0.855 vs. 0.707, respectively, p = 0.031, Jackknife test) when glaucoma suspects were discriminated from healthy, while no significant difference was found when confirmed glaucoma eyes were discriminated from healthy eyes.This new method has the potential to improve early detection of glaucomatous damage.

View Article: PubMed Central - PubMed

Affiliation: University of Pittsburgh Medical Center Eye Center, Eye and Ear Institute, Ophthalmology and Visual Science Research Center, Department of Ophthalmology, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania, USA.

ABSTRACT

Purpose: To develop a new three-dimensional (3D) spectral-domain optical coherence tomography (SD-OCT) data analysis method using a machine learning technique based on variable-size super pixel segmentation that efficiently utilizes full 3D dataset to improve the discrimination between early glaucomatous and healthy eyes.

Methods: 192 eyes of 96 subjects (44 healthy, 59 glaucoma suspect and 89 glaucomatous eyes) were scanned with SD-OCT. Each SD-OCT cube dataset was first converted into 2D feature map based on retinal nerve fiber layer (RNFL) segmentation and then divided into various number of super pixels. Unlike the conventional super pixel having a fixed number of points, this newly developed variable-size super pixel is defined as a cluster of homogeneous adjacent pixels with variable size, shape and number. Features of super pixel map were extracted and used as inputs to machine classifier (LogitBoost adaptive boosting) to automatically identify diseased eyes. For discriminating performance assessment, area under the curve (AUC) of the receiver operating characteristics of the machine classifier outputs were compared with the conventional circumpapillary RNFL (cpRNFL) thickness measurements.

Results: The super pixel analysis showed statistically significantly higher AUC than the cpRNFL (0.855 vs. 0.707, respectively, p = 0.031, Jackknife test) when glaucoma suspects were discriminated from healthy, while no significant difference was found when confirmed glaucoma eyes were discriminated from healthy eyes.

Conclusions: A novel 3D OCT analysis technique performed at least as well as the cpRNFL in glaucoma discrimination and even better at glaucoma suspect discrimination. This new method has the potential to improve early detection of glaucomatous damage.

Show MeSH

Related in: MedlinePlus

Normative database normalization with 46 healthy eyes.(A, B) mean and standard deviation (SD) of retinal nerve fiber layer (RNFL) thickness measurement at each sampling point (A-scan), without normalization. (C, D) mean and SD of RNFL thickness measurement after normalizing individual’s retinal nerve fiber bundle path location to population’s average location. The variations of RNFL thickness were larger at superior temporal and inferior temporal regions (brighter blue in B) because of the population variation of the bundle locations. After aligning the bundle locations and normalizing the RNFL thickness map, the RNFL thickness variations at these two regions were markedly reduced (dark blue in D).
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC3569462&req=5

pone-0055476-g002: Normative database normalization with 46 healthy eyes.(A, B) mean and standard deviation (SD) of retinal nerve fiber layer (RNFL) thickness measurement at each sampling point (A-scan), without normalization. (C, D) mean and SD of RNFL thickness measurement after normalizing individual’s retinal nerve fiber bundle path location to population’s average location. The variations of RNFL thickness were larger at superior temporal and inferior temporal regions (brighter blue in B) because of the population variation of the bundle locations. After aligning the bundle locations and normalizing the RNFL thickness map, the RNFL thickness variations at these two regions were markedly reduced (dark blue in D).

Mentions: A normative database was assembled to measure the RNFL thickness deviation at each sampling point in the 3D dataset. Retinal layer segmentation, which was optimized for 3D SD-OCT dataset [9] was applied on each 3D OCT image to obtain the RNFL thickness and reflectivity at every single sampling point (total 200×200 points). All segmented 3D OCT datasets were visually evaluated to ensure the correct segmentation. Any OCT scan with >8% consecutive B-scan from the total number of B-scans that showed segmentaion error or >12% cumulative segmentation error was excluded from the study. ONH margin was automatically detected on the OCT fundus image using a software program of our own design. [10] Major retinal nerve fiber bundle path at each hemi-field on each RNFL thickness map was automatically detected. Each RNFL thickness map was then normalized by aligning the bundle location at each hemi-field to a reference position (population’s average bundle location) for the comparison with the normative database in order to minimize spatial variability in RNFL thickness, especially at superior temporal and inferior temporal regions (Fig. 2). The RNFL thickness map normalization was processed on the concentrated circles with different radii started from the ONH center. For each subject, the ONH center was first aligned at a reference center point. At each concentrated circle, two average bundle location (superior and inferior) were computed from the normative database. Each subject’s RNFL thickness profile at the given circle was normalized by strenching/shrinking so the subject’s bundle location would coincide with the population average bundle location. The entire RNFL thickness map was normalized by repeating this process at all concentrated circles.


Three-dimensional spectral-domain optical coherence tomography data analysis for glaucoma detection.

Xu J, Ishikawa H, Wollstein G, Bilonick RA, Folio LS, Nadler Z, Kagemann L, Schuman JS - PLoS ONE (2013)

Normative database normalization with 46 healthy eyes.(A, B) mean and standard deviation (SD) of retinal nerve fiber layer (RNFL) thickness measurement at each sampling point (A-scan), without normalization. (C, D) mean and SD of RNFL thickness measurement after normalizing individual’s retinal nerve fiber bundle path location to population’s average location. The variations of RNFL thickness were larger at superior temporal and inferior temporal regions (brighter blue in B) because of the population variation of the bundle locations. After aligning the bundle locations and normalizing the RNFL thickness map, the RNFL thickness variations at these two regions were markedly reduced (dark blue in D).
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC3569462&req=5

pone-0055476-g002: Normative database normalization with 46 healthy eyes.(A, B) mean and standard deviation (SD) of retinal nerve fiber layer (RNFL) thickness measurement at each sampling point (A-scan), without normalization. (C, D) mean and SD of RNFL thickness measurement after normalizing individual’s retinal nerve fiber bundle path location to population’s average location. The variations of RNFL thickness were larger at superior temporal and inferior temporal regions (brighter blue in B) because of the population variation of the bundle locations. After aligning the bundle locations and normalizing the RNFL thickness map, the RNFL thickness variations at these two regions were markedly reduced (dark blue in D).
Mentions: A normative database was assembled to measure the RNFL thickness deviation at each sampling point in the 3D dataset. Retinal layer segmentation, which was optimized for 3D SD-OCT dataset [9] was applied on each 3D OCT image to obtain the RNFL thickness and reflectivity at every single sampling point (total 200×200 points). All segmented 3D OCT datasets were visually evaluated to ensure the correct segmentation. Any OCT scan with >8% consecutive B-scan from the total number of B-scans that showed segmentaion error or >12% cumulative segmentation error was excluded from the study. ONH margin was automatically detected on the OCT fundus image using a software program of our own design. [10] Major retinal nerve fiber bundle path at each hemi-field on each RNFL thickness map was automatically detected. Each RNFL thickness map was then normalized by aligning the bundle location at each hemi-field to a reference position (population’s average bundle location) for the comparison with the normative database in order to minimize spatial variability in RNFL thickness, especially at superior temporal and inferior temporal regions (Fig. 2). The RNFL thickness map normalization was processed on the concentrated circles with different radii started from the ONH center. For each subject, the ONH center was first aligned at a reference center point. At each concentrated circle, two average bundle location (superior and inferior) were computed from the normative database. Each subject’s RNFL thickness profile at the given circle was normalized by strenching/shrinking so the subject’s bundle location would coincide with the population average bundle location. The entire RNFL thickness map was normalized by repeating this process at all concentrated circles.

Bottom Line: Features of super pixel map were extracted and used as inputs to machine classifier (LogitBoost adaptive boosting) to automatically identify diseased eyes.The super pixel analysis showed statistically significantly higher AUC than the cpRNFL (0.855 vs. 0.707, respectively, p = 0.031, Jackknife test) when glaucoma suspects were discriminated from healthy, while no significant difference was found when confirmed glaucoma eyes were discriminated from healthy eyes.This new method has the potential to improve early detection of glaucomatous damage.

View Article: PubMed Central - PubMed

Affiliation: University of Pittsburgh Medical Center Eye Center, Eye and Ear Institute, Ophthalmology and Visual Science Research Center, Department of Ophthalmology, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania, USA.

ABSTRACT

Purpose: To develop a new three-dimensional (3D) spectral-domain optical coherence tomography (SD-OCT) data analysis method using a machine learning technique based on variable-size super pixel segmentation that efficiently utilizes full 3D dataset to improve the discrimination between early glaucomatous and healthy eyes.

Methods: 192 eyes of 96 subjects (44 healthy, 59 glaucoma suspect and 89 glaucomatous eyes) were scanned with SD-OCT. Each SD-OCT cube dataset was first converted into 2D feature map based on retinal nerve fiber layer (RNFL) segmentation and then divided into various number of super pixels. Unlike the conventional super pixel having a fixed number of points, this newly developed variable-size super pixel is defined as a cluster of homogeneous adjacent pixels with variable size, shape and number. Features of super pixel map were extracted and used as inputs to machine classifier (LogitBoost adaptive boosting) to automatically identify diseased eyes. For discriminating performance assessment, area under the curve (AUC) of the receiver operating characteristics of the machine classifier outputs were compared with the conventional circumpapillary RNFL (cpRNFL) thickness measurements.

Results: The super pixel analysis showed statistically significantly higher AUC than the cpRNFL (0.855 vs. 0.707, respectively, p = 0.031, Jackknife test) when glaucoma suspects were discriminated from healthy, while no significant difference was found when confirmed glaucoma eyes were discriminated from healthy eyes.

Conclusions: A novel 3D OCT analysis technique performed at least as well as the cpRNFL in glaucoma discrimination and even better at glaucoma suspect discrimination. This new method has the potential to improve early detection of glaucomatous damage.

Show MeSH
Related in: MedlinePlus