June 2017
Volume 58, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2017
Novel atlas-based score for early glaucoma detection and stratification
Author Affiliations & Notes
  • Fantin Girard
    Polytechnique Montreal, Montreal, Quebec, Canada
  • Farida Cheriet
    Polytechnique Montreal, Montreal, Quebec, Canada
  • Moatez Billah Mekki
    DIAGNOS Medical Algerie, Alger, Algeria
  • Said Yahiaoui
    DIAGNOS Medical Algerie, Alger, Algeria
  • Hadi Chakor
    DIAGNOS Medical Algerie, Alger, Algeria
  • Footnotes
    Commercial Relationships   Fantin Girard, None; Farida Cheriet, None; Moatez Billah Mekki, None; Said Yahiaoui, None; Hadi Chakor, None
  • Footnotes
    Support  CRSNG Grant RDCPJ 484993-15
Investigative Ophthalmology & Visual Science June 2017, Vol.58, 3979. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Fantin Girard, Farida Cheriet, Moatez Billah Mekki, Said Yahiaoui, Hadi Chakor; Novel atlas-based score for early glaucoma detection and stratification. Invest. Ophthalmol. Vis. Sci. 2017;58(8):3979.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : The cup-to-disc ratio (CDR) measurement from the optic disc (OD) appearance in fundus images is one of the key clinical indicator for glaucoma risk assessment. However, the CDR only evaluates the relative sizes of the cup and the OD via their diameters. Local geometric deformations that may be the early signs of glaucoma are not characterized by the CDR. In this work, we propose an automatic novel risk stratification score based on a statistical atlas framework that quantify the presence of the abnormalities of the OD region induced by the early glaucoma condition.

Methods : We build a statistical atlas which describes an average model of the optic disc region, including the vessels, and the variability of local geometric deformations among a healthy population. An atlas-based shape descriptor is then defined as the set of statistical features extracted from the atlas, representing local deviations to the average model and projections on the principal modes of deformation. The novel risk score is then obtained by a linear combination of these statistical features. On a pathological OD with abnormal deformations, the projection on the principal modes and the local deviation will be significantly higher than the healthy atlas variability. The proposed stratification score is evaluated on 350 fundus images including 126 fundus images graded by expert graders as glaucomatous fundus images and 16 at risk OD identified with potential early glaucoma risk condition.

Results : Comparing to the CDR measurement with an area under the ROC curve (AUC) of 89.1%, our novel atlas-based score achieves an AUC of 97.3% which is a significant improvement (p=0.019). On the 16 at risk patients, 14 OD have been identified as significantly different than the healthy atlas showing the capability of the novel score to capture not only glaucomatous patients but also at risk patients.

Conclusions : This study showed that the novel atlas-based score is enough sensitive to detect subtle changes of OD region shape and able to discriminate between healthy and glaucomatous patients. From a practical point of view, this tool is able to assist the clinicians in early glaucoma assessment and may help them for rapid detection of the local changes and for better identification and stratification of the population with glaucoma risk condition.

This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.

 

ROC curve of the CDR and the novel atlas-based score for glaucoma assessment

ROC curve of the CDR and the novel atlas-based score for glaucoma assessment

×
×

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.

×