June 2013
Volume 54, Issue 15
Free
ARVO Annual Meeting Abstract  |   June 2013
Improved 3-Dimensional Retinal Vascular Tree Segmentation and Reconstruction from High-Definition Optical Coherence Tomography
Author Affiliations & Notes
  • Pedro Rodrigues
    CNTM/AIBILI - Association for Innovation and Biomedical Research on Light and Image, Coimbra, Portugal
  • Pedro Guimaraes
    IBILI - Faculty of Medicine, University of Coimbra, Coimbra, Portugal
  • Rui Bernardes
    CNTM/AIBILI - Association for Innovation and Biomedical Research on Light and Image, Coimbra, Portugal
    IBILI - Faculty of Medicine, University of Coimbra, Coimbra, Portugal
  • Pedro Serranho
    IBILI - Faculty of Medicine, University of Coimbra, Coimbra, Portugal
    Mathematics Section, Department of Science and Technology, Open University, Lisbon, Portugal
  • Footnotes
    Commercial Relationships Pedro Rodrigues, None; Pedro Guimaraes, None; Rui Bernardes, None; Pedro Serranho, None
  • Footnotes
    Support None
Investigative Ophthalmology & Visual Science June 2013, Vol.54, 28. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Pedro Rodrigues, Pedro Guimaraes, Rui Bernardes, Pedro Serranho; Improved 3-Dimensional Retinal Vascular Tree Segmentation and Reconstruction from High-Definition Optical Coherence Tomography. Invest. Ophthalmol. Vis. Sci. 2013;54(15):28.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose: To compute the 3D vascular network of the human retina from high-definition spectral-domain optical coherence tomography (OCT).

Methods: The retinal vascular network of the human eye can be directly visualized through several medical imaging modalities, and its application in the assessment of the general health condition has been demonstrated by studies correlating it to retinal, brain (e.g., cognitive ability and cerebrovascular disease), and cardiovascular diseases (e.g., stroke and hypertension). These studies have been performed on 2D fundus images and, consequently, used metrics, such as vessel tortuosity and bifurcation angles, do ignore the depth component. This work aims to compute the 3D vascular structure from OCT volumetric scans by improving the previously proposed approach (Rodrigues et al.; 3D Retinal Vascular Network from OCT data; Invest. Ophthalmol. Vis. Sci. 53: E-Abstract 4099). The algorithm computes an enhanced fundus reference from the OCT volume, followed by the segmentation of the vascular tree through an automatic supervised-learning classification algorithm (support vector machine). A multi-scale wavelet analysis is performed for each A-scan from the vascular tree to find the transition between the hyper-reflectivity and the casted shadow (vessel markers) as an estimate of the depth-wise location of the vessel. Vessel centrelines are thus computed in 3D and the vessel diameter is estimated for each A-scan as the vessel width computed from the 2D fundus reference.

Results: The 3D vascular reconstruction was successfully achieved for OCT scans of healthy controls eyes (N=10) and eyes (N=20) of type 2 diabetic patients (ETDRS levels 10 to 35). Currently, a vascular tree with an extension similar (95% to 110%) to the one provided by colour fundus photography is achieved. In addition, the 3D information is in agreement with the known anatomy of the human retina, with larger vessels found closer to the inner limiting membrane and getting progressively deeper in the retina as vessels divide and become thinner. For the first time, it is possible to visualize crossovers in 3D through a non-invasive imaging modality of the in vivo human retina.

Conclusions: The herewith presented method allows the visualization of the vascular network in 3D and compares favourably to the previous method both in terms of quality and extension of the vascular tree.

Keywords: 549 image processing • 550 imaging/image analysis: clinical • 688 retina  
×
×

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.

×