March 2012
Volume 53, Issue 14
Free
ARVO Annual Meeting Abstract  |   March 2012
A Method for Glaucoma Detection Utilizing ROC Maps Calculated from Three-Dimensional OCT Layer Thicknesses
Author Affiliations & Notes
  • Charles A. Reisman
    Topcon Advanced Biomedical Imaging Laboratory, Topcon Medical Systems, Oakland, New Jersey
  • Qi Yang
    Topcon Advanced Biomedical Imaging Laboratory, Topcon Medical Systems, Oakland, New Jersey
  • Kinpui Chan
    Topcon Advanced Biomedical Imaging Laboratory, Topcon Medical Systems, Oakland, New Jersey
  • Ali S. Raza
    Department of Psychology,
    Columbia University, New York, New York
  • Xian Zhang
    Department of Psychology,
    Columbia University, New York, New York
  • Donald C. Hood
    Department of Psychology,
    Department of Ophthalmology,
    Columbia University, New York, New York
  • Footnotes
    Commercial Relationships  Charles A. Reisman, Topcon Medical Systems, Inc. (E); Qi Yang, Topcon Medical Systems, Inc. (E); Kinpui Chan, Topcon Medical Systems, Inc. (E); Ali S. Raza, None; Xian Zhang, Topcon Medical Systems, Inc. (C); Donald C. Hood, Topcon Medical Systems, Inc. (F, C)
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science March 2012, Vol.53, 679. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Charles A. Reisman, Qi Yang, Kinpui Chan, Ali S. Raza, Xian Zhang, Donald C. Hood; A Method for Glaucoma Detection Utilizing ROC Maps Calculated from Three-Dimensional OCT Layer Thicknesses. Invest. Ophthalmol. Vis. Sci. 2012;53(14):679.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose: : To apply receiver operating characteristic (ROC) analysis to segmented three-dimensional spectral-domain optical coherence tomography (SD-OCT) scans of normal and glaucomatous eyes in order to find optimal conditions for detecting glaucomatous damage.

Methods: : 63 eyes from 63 healthy subjects (controls) (52.4±8.6 yrs) and 122 eyes from 122 patients (57.7±13.4 yrs) were scanned using SD-OCT (Topcon, Inc) with both macular and optic disc fixation. Patient eyes had either normal (suspects) or abnormal (glaucoma) 24-2 visual fields (Zeiss, Inc). The thicknesses of the ganglion cell plus inner plexiform layer (GCL+) and nerve fiber layer (NFL) were determined using an automated algorithm (Yang et al, Opt Exp, 2010). Scans were centered using an automated algorithm. Data sets were excluded based on image quality, clipping, and intra-scan motion criteria. Patients were grouped as G1 (MD≥-3dB, N=77), G2 (-6dB≤MD<-3dB, N=28), and G3 (MD<-6dB, N=17). To generate ROC maps, the 6x6mm 512x128 scan data were reduced to 32x32 NFL (optic disc) and GCL+ (macula) thickness data. The area under the ROC curve (AROC) was calculated independently for each location, creating two-dimensional maps of AROC scores. In optic disc scans, NFL thickness was averaged within annuli of varying diameter and 375 µm in width fitted around the automatically segmented optic nerve head (ONH); and ROC analyses were performed. Traditional circle scan results were simulated by averaging within annuli of fixed 3.4mm diameter.

Results: : Immediately surrounding the ONH, the NFL thickness ROC map for the G1 group exhibited a narrow region with notably higher AROC scores, particularly in the inferior and superior quadrants of the disc. The AROC scores in this region were greater in the G2 and G3 groups and the overall area with high scores was larger. The annulus immediately surrounding the ONH produced AROC scores of 0.913, 0.948, and 0.942 for the G1, G2, and G3 groups, compared to the 3.4mm annulus with respective AROC scores of 0.739, 0.857, and 0.889. For the G1 and G2 groups, the annuli AROCs decreased monotonically with distance from the ONH, while the G3 AROCs increased up to a 375µm distance from the ONH before decreasing monotonically. The AROC scores on the macular GCL+ thickness ROC maps also tended to increase with increased glaucomatous damage (G1 to G3), particularly in the parafoveal region.

Conclusions: : ROC maps are useful for identifying the optimal regions for a test of glaucomatous damage. The mean NFL thickness immediately surrounding the ONH appears to discriminate best between normal and glaucomatous eyes taking into account both sensitivity and specificity.

Keywords: imaging/image analysis: clinical • nerve fiber layer • optic disc 
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×