June 2023
Volume 64, Issue 9
Open Access
ARVO Imaging in the Eye Conference Abstract  |   June 2023
A-Eye: Towards large-scale MRI automated segmentation of the eye
Author Affiliations & Notes
  • Jaime Barranco
    Radiology, Centre Hospitalier Universitaire Vaudois, Lausanne, Vaud, Switzerland
    Universite de Lausanne, Lausanne, Vaud, Switzerland
  • Hamza Kebiri
    Universite de Lausanne, Lausanne, Vaud, Switzerland
    Radiology, Centre Hospitalier Universitaire Vaudois, Lausanne, Vaud, Switzerland
  • Óscar Esteban
    Radiology, Centre Hospitalier Universitaire Vaudois, Lausanne, Vaud, Switzerland
  • Raphael Sznitman
    ARTORG, Universitat Bern Philosophisch-naturwissenschaftliche Fakultat, Bern, Bern, Switzerland
  • Sönke Langner
    Diagnostic and Interventional Radiology, Pediatric and Neuroradiology, Universitatsmedizin Rostock, Rostock, Mecklenburg-Vorpommern, Germany
    Diagnostic Radiology and Neuroradiology, Universitat Greifswald, Greifswald, Mecklenburg-Vorpommern, Germany
  • Oliver Stachs
    Ophthalmology, Universitatsmedizin Rostock, Rostock, Mecklenburg-Vorpommern, Germany
  • Philipp Stachs
    Karlsruher Institut fur Technologie, Karlsruhe, Baden-Württemberg, Germany
  • Benedetta Franceschiello
    Engineering, HES-SO Valais Wallis, Sion, VS, Switzerland
  • Meritxell Bach Cuadra
    Universite de Lausanne, Lausanne, Vaud, Switzerland
    Radiology, Centre Hospitalier Universitaire Vaudois, Lausanne, Vaud, Switzerland
  • Footnotes
    Commercial Relationships   Jaime Barranco, None; Hamza Kebiri, None; Óscar Esteban, None; Raphael Sznitman, None; Sönke Langner, None; Oliver Stachs, None; Philipp Stachs, None; Benedetta Franceschiello, None; Meritxell Bach Cuadra, None
  • Footnotes
    Support  0806-2021 Gelbert Foundation
Investigative Ophthalmology & Visual Science June 2023, Vol.64, PP0010. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Jaime Barranco, Hamza Kebiri, Óscar Esteban, Raphael Sznitman, Sönke Langner, Oliver Stachs, Philipp Stachs, Benedetta Franceschiello, Meritxell Bach Cuadra; A-Eye: Towards large-scale MRI automated segmentation of the eye. Invest. Ophthalmol. Vis. Sci. 2023;64(9):PP0010.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : MRI of the eye (MReye) is raising a lot of interest1, as it provides a comprehensive view of 3D anatomy. This work evaluates an approach for the automated segmentation of eye structures towards the development of a large-scale comprehensive model of the eye.

Methods : Random subset of healthy volunteers (N=1245, age 56±13, 616 females, 594 males) underwent whole-body MRI for the population-based Study of Health in Pomerania (SHIP, Germany)2,3. T1w-images were acquired at 1.5T 1mm3. Manual annotations (9 regions-of-interest, ROIs, Figure 1b) were performed on the right eye on 35 subjects.
Automatic segmentation was achieved by registration4 of a custom template (based on 5 subjects out of those 35 (Figure 1a)), to each remaining subject space (N=30), to determine the bounding box (BB) containing the eye, and then, a second registration within the BB. Atlas labels were projected into the subjects’ space. Agreement metrics5 (Dice Similarity Coefficient (DSC), volume difference, and Hausdorff distance) were reported.

Results : Atlas-based segmentation accurately extracted the 9 ROIs. Figure 1c presents visual label results. Figure 2a presents similarity metrics with respect to the manually annotated dataset. Median DSC values for lens and globe were 0.73, 0.91 respectively. We established new median DSC benchmarks for optic-nerve (0.74), muscles (0.58 to 0.76) and fat (0.67 intraconal and 0.75 extraconal).
Automated segmentation at large-scale dataset. In Figure 2b we report volume estimates per label derived from our automated segmentation on N=1210 grouped by sex.

Conclusions : The proposed method allows to automatically segment lens, vitreous humor, optic nerve, rectus muscles, and fats based on T1w-MRI 1mm3. The performance was benchmarked with manually annotated data. This is the first time feasibility of automated registration-based segmentation and volume estimation on a large-scale dataset is reported, paving the way to create a large-scale MRI-based model of the eye.

This abstract was presented at the 2023 ARVO Imaging in the Eye Conference, held in New Orleans, LA, April 21-22, 2023.

 

a. Custom template construction4 out of 5 subjects. b. Manual annotation on 9 ROI: lens (red), globe (green), optic nerve (dark blue), intraconal (yellow) and extraconal (cyan) fat, lateral (pink), medial (ivory), inferior (purple), and superior (orange) rectus muscle. c. Example of inference from atlas-based segmentation method for the same subject as on b.

a. Custom template construction4 out of 5 subjects. b. Manual annotation on 9 ROI: lens (red), globe (green), optic nerve (dark blue), intraconal (yellow) and extraconal (cyan) fat, lateral (pink), medial (ivory), inferior (purple), and superior (orange) rectus muscle. c. Example of inference from atlas-based segmentation method for the same subject as on b.

 

a. Similarity metrics on N=30. b. Volume estimates in mm3 per structure grouped by sex on N=1210.

a. Similarity metrics on N=30. b. Volume estimates in mm3 per structure grouped by sex on N=1210.

×
×

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.

×