Abstract
To compare the consistency of choroidal hypertransmission (HT) and retinal pigment epithelium (RPE) loss area measurements among three commonly used spectral-domain optical coherence tomography (OCT) devices and assess their robustness for monitoring of geographic atrophy (GA). Patients with GA were imaged during a single visit using the three different OCT: (1) Heidelberg Spectralis; (2) Zeiss Cirrus; (3) Topcon Maestro2. Expert readers manually annotated a total of 120 OCT volumes for RPE loss and HT. Dice similarity coefficients (DSC) were calculated to quantify the spatial overlap between the lesions within each OCT device. Intraclass correlation coefficients (ICC) and Bland-Altman analyses were used to assess inter-device agreement. Spatial overlap between HT and RPE loss ranged from moderate to good, with HT lesions being significantly larger than RPE loss areas across all devices (p < 0.001). Overall agreement was good for HT and RPE loss. Systematic biases emerged, with Cirrus yielding consistently smaller measurements compared to Spectralis or Maestro2. Lesion measurements demonstrated good agreement across all three OCT devices, yet device-dependent differences require caution when comparing data. Accounting for inter-device variability is an essential step toward reliable clinical endpoints and successful integration of automated OCT algorithms into clinical trials and routine AMD care.