May 2005
Volume 46, Issue 13
ARVO Annual Meeting Abstract  |   May 2005
A Robust Procedure for Vessel Tracking in Retinal Images
Author Affiliations & Notes
  • A. Ruggeri
    Dept. Information Engineering, University of Padova, Padova, Italy
  • E. Grisan
    Dept. Information Engineering, University of Padova, Padova, Italy
  • A. Giani
    Dept. Information Engineering, University of Padova, Padova, Italy
  • Footnotes
    Commercial Relationships  A. Ruggeri, Nidek Technologies C; E. Grisan, Nidek Technologies F; A. Giani, None.
  • Footnotes
    Support  None.
Investigative Ophthalmology & Visual Science May 2005, Vol.46, 1557. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      A. Ruggeri, E. Grisan, A. Giani; A Robust Procedure for Vessel Tracking in Retinal Images . Invest. Ophthalmol. Vis. Sci. 2005;46(13):1557.

      Download citation file:

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

  • Supplements

We propose a new, fully automatic procedure for the extraction of the vascular structure and related features in retinal images.


The procedure is based on a sparse tracking technique. Starting from a set of seed points, spread all over the image, the tracking procedure moves along the vessel by analyzing subsequent vessel cross sections (small green segments in figure). In each cross section, points belonging to the vessel are found by means of a fuzzy C–means classifier, and, from these points, the algorithm extracts the vessel center, caliber and direction. When tracking stops because of a critical area (e.g. a vessel bifurcation or crossing, or an area with low contrast,), an innovative ''bubble technique'' module is run. It is able to draw and analyze multiple circular scan lines around the critical point (see figure), allowing the exploration of the region around it and thus to resume vessel tracking beyond the critical area. After the tracking step is completed, a search is run to recognize and eliminate possible false vessels, e.g. choroidal vessel segments identified as retinal vessels. Finally, to reconstruct the course of whole vessels, the sparse vessel segments are linked by a connection algorithm, taking into account similarity of caliber, direction and color of the pair of facing segments under examination. Bifurcations and crossings are then identified by analyzing specific features of vessel segments.


On a data set containing sixty 50–degrees color images, acquired with a TRC–50 fundus camera, (Topcon Co., Japan), from both normal and pathological subjects, the proposed algorithm was able to recognize on average 92% (min 82%, max 98%) of the manually tracked vessel structure, extracting 88% (min 67%, max 98%) of the bifurcations and 57% (min 24%, max 94%) of the crossings.


In our data set of images, the procedure proved effective to extract a large portion of the vascular structure and a good percentage of bifurcations and crossings. This information was used to derive in each image the position of the optic disc and other parameters of clinical significance, such as vessel tortuosity and generalized arterial narrowing.



Keywords: imaging/image analysis: clinical • 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.