Empirical Distortion Correction for Posterior Segment SD-OCT Images in Small Animals

Abstract

Purpose : Develop an empirical distortion correction approach for commercial posterior segment SD-OCT devices that use a fan scanning pattern but otherwise undisclosed optical details.

Methods : Based on the assumption that the A-Scan rays that make up a single B-scan converge at one pivot point in space, the image space is mapped to the object space by a nonlinear transformation defined by three unknown variables: axial and angular scaling factors, and pivot point location. We used two types of data sets to empirically estimate these parameters in 3 tree shrew eyes: (i) MRI images of each tree shrew eye were obtained and the curvature of the posterior ocular coats were manually delineated. (ii) Glass beads of known diameter (100±2.7 μm) were implanted on top of or into the retina. The Spectralis OCT2 (Heidelberg Engineering) was used to acquire B-scans through the center of the implanted beads while varying camera settings (camera position, scan focus, and reference arm position [RAP]). The glass bead diameter and RPE were manually delineated from 120 B-Scans. The unknown transformation parameters were identified by matching the mapped bead diameter and RPE curvature for each B-scan to the known diameter and MRI-based curvature, respectively. The proposed method was validated against distorted images of the optic nerve head (ONH) by imaging the same ONH at baseline position and after moving the camera 5 mm anterior or posterior.

Results : The algorithm used was able to restore the bead diameter and RPE curvature in the optically distorted OCT images. A clear non-linear correlation between the angular scaling factor and the pivot point location was identified (R²: 0.997). The axial scaling factor was independent of the device settings. The angular scaling factor was correlated with two scan parameters: scan focus and RAP (R²: 0.687). Based on the identified correlations, anatomically corrected ONH images were restored from the distorted OCT scans.

Conclusions : No commercial OCT device provides anatomically correct images. We have proposed an empirical approach to compensate for non-linear distortions in SD-OCT images from commercial devices that use fan scanning. Our method does not require knowledge of the device’s optical system.

This is a 2020 ARVO Annual Meeting abstract.