Use of fundus image registration and flicker for tracking choroidal tumor changes

Abstract

Purpose : Fundus photography is important in documenting and assessing the progression of uveal tumors and their potential therapeutic ocular complications. The purpose of this retrospective study is to investigate the use of registration and flicker of serial fundus images to detect changes in choroidal tumors that may indicate transformation to malignancy or development of therapy-induced retinopathy.

Methods : Twenty-two eyes from 21 subjects with progression of choroidal tumors or therapeutic complications (e.g., radiation retinopathy) were imaged during multiple visits (minimum of 3 and maximum of 8 visits during follow-up period) using CLARUS^TM 500 (ZEISS, Dublin, CA) with true color widefield imaging. A custom registration algorithm was used to register images from consecutive follow-up visits in addition to those from baseline to the latest visit. All the registered pairs had a confidence metric of equal to or greater than 0.7. An animated flicker image with 500ms delay was created for each of the registered pairs.

An optometrist graded the side-by-side unregistered images, side-by-side registered images, and flickered fundus images from consecutive patient visits for change in intraocular tumor. Changes were graded on a scale of three: 0 for no change, 1 for possible change, and 2 for change (progression or regression).

Results : Figure 1 shows the results of the analyses. The changes in tumor or therapeutic complications were easier to track in flicker images than the unregistered and registered images in both consecutive visits and baseline to the latest visit. Registered images were marginally better in tracking changes than the unregistered images.

Conclusions : The advantage of flicker in detecting subtle changes in choroidal tumors and therapy-induced retinopathy, using registered fundus images from consecutive patient visits, was demonstrated. Further studies are needed with multiple graders and a more diverse data set for detailed analysis.

This is a 2020 ARVO Annual Meeting abstract.

 

Figure 1. Summary of results.

 

Figure 2. (a) Both unregistered (top left) and registered (bottom left) were graded as "possible change" in choroidal melanoma and radiation retinopathy post plaque radiation. Their corresponding animated flicker was graded as "change" due to observed disc hemorrhage and nerve fiber layer infarction along with microaneurysms. (b) All image presentations were graded as "change" due to observable change in tumor diameter.