June 2023
Volume 64, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2023
Anterior segment axial motion correction using the cornea surface
Author Affiliations & Notes
  • Gwen Musial
    Carl Zeiss Meditec, Inc., Dublin, California, United States
  • Homayoun Bagherinia
    Carl Zeiss Meditec, Inc., Dublin, California, United States
  • Zahra Nafar
    Carl Zeiss Meditec, Inc., Dublin, California, United States
  • Footnotes
    Commercial Relationships   Gwen Musial Carl Zeiss Meditec, Inc., Code E (Employment); Homayoun Bagherinia Carl Zeiss Meditec, Inc., Code E (Employment); Zahra Nafar Carl Zeiss Meditec, Inc., Code E (Employment)
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science June 2023, Vol.64, 3396. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Gwen Musial, Homayoun Bagherinia, Zahra Nafar; Anterior segment axial motion correction using the cornea surface. Invest. Ophthalmol. Vis. Sci. 2023;64(8):3396.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Imaging of the anterior segment (AS) using optical coherence tomography (OCT) is useful for surgical planning, post-operative monitoring, and disease diagnosis. However, axial motion during OCT acquisition degrades image scan quality leading to ambiguous clinical interpretation and even erroneous diagnosis. Correcting axial motion during AS imaging is challenging due to low contrast in the cornea compared to the iris and changing pupil sizes during scan acquisition. Therefore, we developed an axial motion correction algorithm using deep learning (DL) segmentation of the cornea surface as a reference to improve image quality.

Methods : Axial motion estimation and correction is based on a pair of reference scans that represent motion free sparse scans. Matching the cornea surface between reference scans and corresponding slow B-scans determines the motion estimation for each B-scan in the cube.
The DL-based model segments the cornea surface of the reference scans and B-scans. A polynomial model fit using RANSAC models the cornea surface of the reference segmentation. Then, the axial motion compensation values for each B-scan in the fast direction are calculated as the difference between the model fit and the segmentation of the corresponding cross section of the volume segmentation points.
Performance of the algorithm was evaluated using 60 AS Angio 6 mm x 6 mm x 6 mm PLEX® Elite 9000 (ZEISS, Dublin, CA) swept-source OCT scans from 13 healthy subjects by manual grading. A grader reviewed the slow B-scans and the corresponding reference images and rated the level of axial motion correction quality for each segmented surface as being 3: acceptable, 2: partially acceptable, or 1: failed.

Results : Figure 1 shows examples for two reference images and corresponding slow B-scans before and after axial motion correction. The success rating across the entire data set is 77% grade 3, 16% grade 2, and 7% grade 1.

Conclusions : We developed an algorithm for AS axial motion correction using corneal surface segmentation. Assessment of the algorithm shows that improved axial motion correction results in more acceptable image quality with 93% at least partially acceptable. Implementation of the algorithm will assist clinical assessment of the entire anterior chamber.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.

 

Figure 1: Two reference scans (at 20% and 80% of the fast scan direction) and corresponding slow B-scans sampled from the cube before and after correction with quality grade 3.

Figure 1: Two reference scans (at 20% and 80% of the fast scan direction) and corresponding slow B-scans sampled from the cube before and after correction with quality grade 3.

×
×

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.

×