June 2015
Volume 56, Issue 7
Free
ARVO Annual Meeting Abstract  |   June 2015
Feasibility of automated interface segmentation of Cirrus HD-OCT data in normal and mild non proliferative diabetic retinopathy eyes
Author Affiliations & Notes
  • Torcato Santos
    Centre of New Technologies for Medicine, Association for Innovation and Research on Light and Image, Coimbra, Portugal
  • Antonio Correia
    Centre of New Technologies for Medicine, Association for Innovation and Research on Light and Image, Coimbra, Portugal
  • Catarina A Neves
    Coimbra Ophthalmology Reading Centre, Association for Innovation and Biomedical Research on Light and Image, Coimbra, Portugal
  • Christian Schwartz
    Coimbra Ophthalmology Reading Centre, Association for Innovation and Biomedical Research on Light and Image, Coimbra, Portugal
  • Telmo Miranda
    Centre of New Technologies for Medicine, Association for Innovation and Research on Light and Image, Coimbra, Portugal
  • Ana Rita Santos
    Coimbra Ophthalmology Reading Centre, Association for Innovation and Biomedical Research on Light and Image, Coimbra, Portugal
  • Jose G Cunha-Vaz
    Association for Innovation and Biomedical Research on Light and Image, Coimbra, Portugal
  • Footnotes
    Commercial Relationships Torcato Santos, None; Antonio Correia, None; Catarina Neves, None; Christian Schwartz, None; Telmo Miranda, None; Ana Rita Santos, None; Jose Cunha-Vaz, Carl Zeiss Meditec (C)
  • Footnotes
    Support None
Investigative Ophthalmology & Visual Science June 2015, Vol.56, 5953. doi:https://doi.org/
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Torcato Santos, Antonio Correia, Catarina A Neves, Christian Schwartz, Telmo Miranda, Ana Rita Santos, Jose G Cunha-Vaz; Feasibility of automated interface segmentation of Cirrus HD-OCT data in normal and mild non proliferative diabetic retinopathy eyes. Invest. Ophthalmol. Vis. Sci. 2015;56(7 ):5953. doi: https://doi.org/.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose: Implementation and validation of a graph based segmentation algorithm on Optical Coherence Tomography (OCT) acquired data from eyes with preserved retinal structure, using normal control and mild Non Proliferative Diabetic Retinopathy (NPDR) populations.

Methods: The graph theory segmentation algorithm in [1] was implemented to automatically identify 8 retinal interfaces namely Vitreous to Inner Limiting Membrane (ILM), Retinal Nerve Fiber to Ganglion Cell Layers, Inner Plexiform (IPL) to Inner Nuclear Layers (INL), INL to Outer Plexiform Layer (OPL), OPL to Outer Nuclear Layer (ONL), ONL to Inner Segment (IS), Outer Segment to Retinal Pigment Epithelium (RPE) and RPE to Choroid. The algorithm was applied on Cirrus HD-OCT (Carl Zeiss Meditec, Dublin, CA, USA) acquired data from 86 eyes of 47 healthy volunteers aged from 23 to 55 years (m±SD: 38.30±9.10) and from 72 eyes of 72 NPDR patients, including subclinical and clinical macular edema (DRCRnet criteria), aged from 39 to 75 years (m±SD: 60.01±8.70).

Results: Root mean square errors (RMSE) between automatic and human grader segmentations for healthy volunteers and for NPDR patients are of the same order of magnitude. Larger RMSE were found at the OS/RPE interface ranging from 3.97 to 16.61 (m±SD: 6.92±2.21) [µm] and from 0.66 to 17.76 (m±SD: 4.24±2.69) [µm] for healthy subjects and for NPDR patients respectively. Interfaces at the Central Sub Field have larger RMSE relatively to other ETDRS areas, ranging from 2.58 to 9.02 [µm] and from 0.79 to 8.50 [µm] for healthy subjects and for NPDR patients respectively. Lower RMSE were found at the outer rings for all interfaces with the exception of the vitreous/ILM and ONL/IS for healthy volunteers and vitreous/ILM and IPL/INL for NPDR patients.

Conclusions: RMSE between automatic and human grader segmentations are close to the device 5 µm axial resolution in tissue which makes this segmentation algorithm well suited for detecting individual retinal layer changes in situations where the retinal structure is well preserved, such as mild NPDR, contributing to identify the relative role of the different retinal cells in the retinopathy development.<br /> <br /> [1] Li K., Wu X., Chen D., Sonka M., “Optimal Surface Segmentation in Volumetric Images - A Graph-Theoretic Approach”, IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 28, no. 1, pp. 119-134, January 2006.

×
×

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.

×