Abstract
Purpose :
Spectral Domain Optical Coherence Tomography (SD-OCT) is essential in clinical ophthalmology for providing detailed cross-sectional images and measurements of the retina. Using advanced computational methods, we evaluated retinal thickness measurements from older and newer Spectralis SD-OCT devices.
Methods :
In this study, we evaluated the reproducibility and quality of retinal thickness measurements from two OCT machines using data from 15 patients. We calculated average pairwise Pearson’s correlations from repeated measurements and conducted Bland-Altman analysis for additional validation. Multidimensional scaling (MDS) and principal component analysis (PCA) were used to visualize key attributes of the different retinas measured. Signal-to-noise Ratio (SNR) and Contrast-to-Noise (CNR) were computed as quality metrics. Variance analyses were conducted to identify retinal regions with inconsistent measurements, as well as retinal regions with the highest intrinsic variation in thickness among different patients. Finally, as proof-of-concept, a support vector machine (SVM) classifier was trained to distinguish between right and left eye measurements.
Results :
Both OCT machines showed high reproducibility in total retinal thickness measurements with a median APPC of 99.8%. PCA validated consistent measurements with each patients’ repeated tests forming tight clusters, and identified two larger clusters corresponding to the sidedness of the eye. Spatial variance analysis revealed that peripheral regions of the retina were less consistently measured compared to central regions of the retina. Moreover, compared to nasal regions, the temporal regions of the retina varied more in total thickness across different patients. The machine learning classifier accurately categorized the sidedness of the eye (AUROC = 1.0, accuracy = 100%).
Conclusions :
This research confirms that Spectralis SD-OCT instruments, both old and new, reliably measure retinal thickness. Automated classification's high precision suggests improved diagnostic capabilities, while observed spatial variances point to opportunities for advancing ophthalmic imaging technology.
This abstract was presented at the 2024 ARVO Imaging in the Eye Conference, held in Seattle, WA, May 4, 2024.