Retinal image registration approach in optical coherence tomography and fluorescence imaging

Otesteanu, Corin F.; Robledo, Lucio; Zinkernagel, Martin; Sznitman, Raphael; Márquez Neila, Pablo (6 January 2023). Retinal image registration approach in optical coherence tomography and fluorescence imaging (In Press). Investigative ophthalmology & visual science. Association for Research in Vision and Ophthalmology

	Text ARVOShortVersionSubmitted_-_Retinal_image_registration_approach_in_optical_coherence_tomography_and_fluorescence_imaging.pdf - Accepted Version Restricted to registered users only Available under License Publisher holds Copyright. Download (1MB)
	Text ARVOShortVersionSubmitted_-_Retinal_image_registration_approach_in_optical_coherence_tomography_and_fluorescence_imaging.pdf - Submitted Version Restricted to registered users only Available under License Publisher holds Copyright. Download (1MB)

Purpose: To investigate the feasibility of an in-house built device for acquiring optical coherence tomography (OCT) and fluorescence images simultaneously for macular imaging. Registration of these modalities is necessary to correct motion artifacts that degrade image quality.

Methods: OCT and fluorescent volumes were consecutively acquired from two healthy volunteers and of a model eye. Data were acquired in the following axes order: depth (0z), lateral (0x), and azimuthal (0y). An additional 2D slice was acquired in the depth-azimuthal plane.
First, 2D Bscans (x0z slices) were registered in every 3D volume–intra-volume registration. Gradient descent optimization with mutual information as a cost function was used to solve this 2D rigid registration problem. For registration validation purposes, a 2D slice that was acquired in the y0z plane was used as ground truth. Results were compared to this using the mean squared error metric (MSE).
Second, the 3D volumes of a patient were registered with each other–inter volume registration. A similar approach to that described was used to solve this as a 3D rigid registration problem.

Results: A decrease in MSE of 15.5% (±17.4% std) was obtained after the Bscan registration. Figure 1a,b show a y0z slice pre- and post-registration of an acquisition with various motion pulses. The peaks in the retina from the pre-registration images have been corrected in the post-registered image. The 3D registered volumes were averaged, and the corresponding enface image was computed. A sharper optic nerve head (ONH) image was obtained in the post-registration (Fig. 1c). Figure 2a shows a fluoresce ONH image centered on the fovea, visibly sharper post-registration. A side-by-side comparison of a model eye fluorescence and enface images after registration shows a good correspondence of anatomical structures (Fig. 2b).

Conclusions: We showed that OCT and fluorescence retinal image errors due to motion artifacts can be improved using the presented registration approach. A 15.5% improvement was achieved in MSE. Visually the resulting images were sharper or with fewer artifacts.

Interest & Impact

Downloads

9 since deposited on 07 Feb 2023
2 in the past 12 months

Citations

5 Citations in Web of Science ®
6 Citations in Scopus

Search

in Google Scholar™

Services

Actions (login required)

Edit item

Actions (login required)

Edit item