August 2019
Volume 60, Issue 11
Free
ARVO Imaging in the Eye Conference Abstract  |   August 2019
Segmentation of inner microlayers of abnormal cornea in Optical Coherence Tomography images using graph segmentation
Author Affiliations & Notes
  • Amr Elsawy
    Ophthalmology, Bascom Palmer Eye Institute, Miami, Florida, United States
    Electrical and Computer Engineering, University of Miami, Coral Gables, Florida, United States
  • Taher Kamel Eleiwa
    Ophthalmology, Bascom Palmer Eye Institute, Miami, Florida, United States
  • Mariam Raheem
    Ophthalmology, Bascom Palmer Eye Institute, Miami, Florida, United States
  • Giovanni Gregori
    Ophthalmology, Bascom Palmer Eye Institute, Miami, Florida, United States
  • Mohamed Abdel-Mottaleb
    Electrical and Computer Engineering, University of Miami, Coral Gables, Florida, United States
  • Mohamed Abou Shousha
    Ophthalmology, Bascom Palmer Eye Institute, Miami, Florida, United States
    Electrical and Computer Engineering, University of Miami, Coral Gables, Florida, United States
  • Footnotes
    Commercial Relationships   Amr Elsawy, None; Taher Eleiwa, None; Mariam Raheem, None; Giovanni Gregori, None; Mohamed Abdel-Mottaleb, None; Mohamed Abou Shousha, NEI core center grant to the University of Miami (P30 EY014801) (P), NEI K23 award (K23EY026118) (F), Research to Prevent Blindness (RPB) (F)
  • Footnotes
    Support  NEI K23 award (K23EY026118), NEI core center grant to the University of Miami (P30 EY014801), and Research to Prevent Blindness (RPB)
Investigative Ophthalmology & Visual Science August 2019, Vol.60, 024. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Amr Elsawy, Taher Kamel Eleiwa, Mariam Raheem, Giovanni Gregori, Mohamed Abdel-Mottaleb, Mohamed Abou Shousha; Segmentation of inner microlayers of abnormal cornea in Optical Coherence Tomography images using graph segmentation. Invest. Ophthalmol. Vis. Sci. 2019;60(11):024.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : To report on a graph-based automated segmentation method to segment the epithelium, Bowmans (BW), Stroma (ST) and Descemets (DM) microlayers of abnormal cornea using high definition optical coherence tomography (HD-OCT) images.

Methods : Fifteen eyes were imaged using HD-OCT (Envisu R2210, Bioptigen, Buffalo Grove, IL, USA). These eyes had different diagnoses: Dry eye (2 eyes), Keratoconus (5 eyes), Fuchs Dystrophy (3 eyes), and controls (5 eyes). The 15 images were manually segmented by two expert operators. The automated segmentation algorithm segmented the Epithelium (EP) and the Endothelium (EN) microlayers and used them to segment the inner microlayers. A flat-EP image was obtained, and its vertical gradient was computed. A directed graph was used to segment the BW and ST as the two hyper-reflective microlayers after the EP. A flat-EN image was obtained and its vertical gradient was computed. A directed graph was used to segment the DM as the hyper-reflective microlayer before the EN. In both graphs, the edge energy included gradient energy and directional energy. Some of the segmentation results are shown in Fig. 1.

Results : The epithelium, Bowmans, Stroma and Descemets microlayers of those abnormal corneas were successfully segmented using the graph method in all images. The inter-operator error between the two manual operators was 1.75±1.54 pixels for the epithelium, Bowmans, 1.73±1.43 for the Stroma and 3.10±2.28 for the Descemets (mean 2.17±1.88 pixels). The mean segmentation difference between the graph-based segmentation and the manual operators was 1.69±1.90 pixels for the epithelium, Bowmans, 2.47±3.15 pixels for the Stroma and 4.36±3.21 for the Descemets (mean 2.77±2.99 pixels). The segmentation results were comparable to the manual segmentation (see Fig. 2). The mean segmentation time for all microlayers in one image was 4.27±0.65 seconds and the segmentation had zero intra-operator error.

Conclusions : The proposed graph-based segmentation method, for segmenting corneal microlayers using HD-OCT images, was objectively comparable to the manual segmentation for abnormal and healthy corneas and it is significantly faster and repeatable.

This abstract was presented at the 2019 ARVO Imaging in the Eye Conference, held in Vancouver, Canada, April 26-27, 2019.

 

Fig. 1 Examples of segmentation of inner corneal microlayers

Fig. 1 Examples of segmentation of inner corneal microlayers

 

Fig. 2 Segmentation difference comparison

Fig. 2 Segmentation difference comparison

×
×

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.

×