Weighted Feature Matching registration for tracking choroidal tumors using flicker presentation

Abstract

Purpose : Fundus imaging is essential in documenting choroidal tumors and the comparison of images between visits is critical to assessing their progression and therapeutic complications. When comparing images, it is important that the region of interest is aligned and registered to accurately evaluate change. The purpose of this study is to investigate the use of Weighted Feature Matching (WFM) registration with flicker presentation of fundus images to detect changes in choroidal tumors.

Methods : In this retrospective study, 22 eyes from 21 subjects with choroidal tumors or therapeutic complications such as radiation retinopathy were imaged over consecutive visits (154 ± 105.7 days between visits) using CLARUS^TM 500 (ZEISS, Dublin, CA) with true color widefield imaging.

WFM registration is a new algorithm to improve peripheral retinal registration with minimum reduction to performance in the center of the image. An animated flicker image with 500ms delay was created for each of the registered pairs.

A clinical grader evaluated side-by-side unregistered, WFM registered, and WFM flickered images of consecutive visits for change in intraocular tumors. Changes were graded as: 0 no change, 1 possible change, and 2 change (progression or regression).

Results : The results of the analyses are shown in figure 1. The lesion changes were easier to track in flicker images than the unregistered and registered images over consecutive visits. Registered images displayed side-by-side were better in tracking changes than the unregistered images displayed side-by-side. The use of WFM method in registration produced comparable clinical performance to previously published literature (Su et al. IOVS 2020; 61(7):3640).

Conclusions : In this project, WFM registered flicker images performed better in detecting subtle changes in choroidal tumors than the registered and unregistered side-by-side images. Future studies with include multiple graders and a larger data set with different pathologies.

This is a 2021 Imaging in the Eye Conference abstract.

 

Figure 1. Analyses of grades between visits

 

Figure 2. A) Unregistered (top left) and registered (bottom left) were graded as “no change“. The flicker presentation was graded as "change" due to observed elevation of the lesion. B) Unregistered were graded as “no change”, WFM registered and flicker were graded as “change” due to advancement of superior border. C) All image presentations were graded as "change" due to increase in tumor diameter.