Purpose:
To develop and evaluate a new algorithm to estimate the relative location of StratusOCTTM scan circles (SCs) for retinal nerve fibre layer thickness (RNFLT) measurement.
Methods:
The SC relative location inference algorithm has four components: 1) the retinal pigment epithelium (RPE) is detected, and the RPE height modulation is corrected by realigning the RPE; 2) candidate vessels are detected utilizing a linear perceptron and feature detection using the image integral centered along the image patch around the RPE; 3) the detected vessels in sequential images are then matched by the novel adaptation of a dynamic programming technique commonly used in bioinformatics to align nucleotide sequences; 4) assuming that the SC movement includes shifts, rotations and incline, and that the vessels within the SC images region are straight lines, the algorithm infers relative displacement using scaled conjugate gradients. The new algorithm was evaluated by examining the impact on the reproducibility of mean quadrant RNFLT values (coefficient of variation; CV), using large samples of repeated scans on 20 subjects. The algorithm was used to identify groups of scans that were most closely aligned.
Results:
The relative position (Fig) of SCs identified by the new algorithm is directly related to the reproducibility of RNFLT measurements: mean CV was significantly reduced (6.9% to 4.1% for Temporal, 5.4% to 2.5% for Superior, 10.1% to 5.0% for Nasal, 6.1% to 3.0% for Inferior; P<0.05) for estimates of mean quadrant RNFLT in images where the SC centers were closely matched compared to those that were not.
Conclusions:
The relative location of SCs as estimated by a novel algorithm is related to the reproducibility of RNFLT measurements from the StratusOCTTM. Furthermore, when combined with improved segmentation algorithms these methods can facilitate true image 'averaging', where increased number of scans at the point of acquisition can be used to compose a wider 3D view of RNFL structure (Fig); these images might be useful for detecting structural progression in glaucoma.
Keywords: imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • imaging/image analysis: clinical • retina