June 2021
Volume 62, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2021
Weighted feature point matching for registration of widefield fundus images
Author Affiliations & Notes
  • Niranchana Manivannan
    Carl Zeiss Meditec, Inc., Dublin, California, United States
  • Susan Su
    Carl Zeiss Meditec, Inc., Dublin, California, United States
  • Archana Kolli
    Carl Zeiss Meditec, Inc., Dublin, California, United States
  • Angelina Covita
    Carl Zeiss Meditec, Inc., Dublin, California, United States
  • Sandor Ferenczy
    Ocular Oncology Service, Wills Eye Hospital, Philadelphia, Pennsylvania, United States
  • Carol Shields
    Ocular Oncology Service, Wills Eye Hospital, Philadelphia, Pennsylvania, United States
  • Mary K Durbin
    Carl Zeiss Meditec, Inc., Dublin, California, United States
  • Footnotes
    Commercial Relationships   Niranchana Manivannan, Carl Zeiss Meditec, Inc., Dublin, CA, United States (E); Susan Su, Carl Zeiss Meditec, Inc., Dublin, CA, United States (C); Archana Kolli, Carl Zeiss Meditec, Inc., Dublin, CA, United States (E); Angelina Covita, Carl Zeiss Meditec, Inc., Dublin, CA, United States (E); Sandor Ferenczy, None; Carol Shields, None; Mary Durbin, Carl Zeiss Meditec, Inc., Dublin, CA, United States (E)
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science June 2021, Vol.62, 1794. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Niranchana Manivannan, Susan Su, Archana Kolli, Angelina Covita, Sandor Ferenczy, Carol Shields, Mary K Durbin; Weighted feature point matching for registration of widefield fundus images. Invest. Ophthalmol. Vis. Sci. 2021;62(8):1794.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : As most features in the retina are found in the posterior pole, including the optical nerve head (ONH), the performance of registration algorithms when applied to widefield fundus images is generally sub-optimal in the periphery. In this research, we propose a weighted feature matching (WFM) to improve registration over the entire field of view of the widefield image.

Methods : Figure 1 shows the steps of the proposed algorithm. Both fixed and moving image pairs for registration are channel separated. Green channel is used for ONH detection [1]. Cornerness map from multi-scale Harris corner detection is multiplied with wi before thresholding. If the fixed image is of size m x n and di is pixel distance from either the ONH (if detected) or from the center of the image (if ONH is not detected), the weighting factor is wi= di / (m*n). Weighted vector distances between histogram of oriented gradients (HOG) descriptors (16 × 16 patches) are used for the feature point matching. Two registered images with and without WFM are generated for comparison.
Baseline and follow up images from 44 subjects (22 healthy and 22 with choroidal tumors, therapy-induced retinopathy, diabetic retinopathy and glaucoma) are acquired using CLARUSTM 500 (ZEISS, Dublin, CA) . For each pair of images, cpselect function in MATLAB is used to label 10 landmarks/points per image (5 in the center and 5 in the periphery). The angular distance between the points in the registered images should be zero degrees for perfect registration. The mean angular distance with and without WFM are calculated for the quantitative evaluation.

Results : The mean angular distance between the points in the registered images with and without WFM are 0.60±0.31°, 0.73±0.38° (all points), 0.56±0.25°, 0.54±0.28° (central points) and 0.63±0.35, 0.92±0.38° (peripheral points). With WFM, the overall performance of the registration is improved by 17.8% while the performance in the center of the image is decreased by 2.2%. Figure 2 shows the registered image with and without WFM.

Conclusions : The proposed registration with WFM improves the overall registration of the widefield fundus image by improving the performance in the periphery with minimum reduction to the performance in the center of the image.

Reference:
[1] Meng et al. IOVS 2020; 61(9): Abstract PB0060.

This is a 2021 ARVO Annual Meeting abstract.

 

Figure 1. Flowchart of the proposed algorithm

Figure 1. Flowchart of the proposed algorithm

 

Figure 2. Unregistered images (a,b), without WFM (c,d) and with WFM (e,f)

Figure 2. Unregistered images (a,b), without WFM (c,d) and with WFM (e,f)

×
×

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.

×