Computational analysis of reproducibility and image quality of 2011 and 2017 models of Heidelberg Spectralis SD-OCT machines

Abstract

Purpose

Evaluate the integrity, reproducibility, and image quality of total retinal thickness (TRT) measurements between two generations of Spectralis Spectral Domain Optical Coherence Tomography (SD-OCT) instruments (Old OCT: 2011, New OCT: 2017).

Design

Prospective cohort study evaluating TRT measurements across two visits.

Subjects and participants

Fourteen healthy individuals (28 eyes, age range: 22-54 years) underwent TRT measurements using both Old and New OCT models, with each eye receiving three consecutive scans per visit.

Methods and testing

TRT measurements were performed using the Posterior Pole Algorithm (PPA) and Early Treatment Diabetic Retinopathy Study (ETDRS) grid protocols. Reproducibility was evaluated using Average Pairwise Pearson Correlation (APPC), while image quality was measured by Signal-to-Noise Ratio (SNR) and Contrast-to-Noise Ratio (CNR). Agreement between the devices was analyzed through Bland-Altman plots, and spatial variability was visualized using heatmaps. The dimensionality reduction techniques Principal Component Analysis (PCA) and Multidimensional Scaling (MDS) were employed to explore data patterns.

Main outcome measures

Reproducibility of TRT measurements, image quality, and the degree of agreement between the two OCT models.

Results

Both the Old and New OCT models demonstrated high reproducibility (APPC: 0.995-0.998). While there was not a statistically significance difference in reproducibility between the OCT models, image quality analysis revealed superior SNR and CNR values for the New OCT in the left eye only, with significant improvements noted (CNR: p = 0.0040 at Visit 2; SNR: p = 0.0383 at Visit 1). Bland-Altman analysis confirmed strong agreement, with minimal mean differences and narrower limits of agreement for the New OCT. Heatmap analysis indicated greater inter-patient variability in the nasal retinal regions, while intra-patient variability was consistently low (<1%) across both devices. PCA and MDS plots affirmed the reproducibility of measurements.

Conclusions

Both SD-OCT models provide reliable and consistent TRT measurements, with the New OCT offering marginally enhanced image quality. However, the reproducibility of the New OCT does not significantly outperform the Old OCT, supporting the use of both devices for accurate TRT assessment in clinical settings. Further studies may be required to evaluate these findings in pathological conditions.