Performance of drusen detection by spectral-domain optical coherence tomography

Area of Science:

• Ophthalmology
• Medical Imaging
• Biomedical Engineering

Background:

• Age-related macular degeneration (AMD) is a leading cause of vision loss.
• Early AMD is characterized by the presence of drusen.
• Accurate detection and quantification of drusen are crucial for monitoring AMD progression.

Purpose of the Study:

• To assess the performance of automated analyses in spectral-domain optical coherence tomography (SD-OCT) devices for drusen identification.
• To compare the accuracy of automated drusen detection across three different SD-OCT devices.
• To evaluate the limitations of automated segmentation algorithms in quantifying drusen in early AMD.

Main Methods:

• Twelve eyes of AMD patients (AREDS 2 and 3) with a mean of 113 drusen were scanned using three SD-OCT devices (Cirrus, 3DOCT-1000, Spectralis).
• Automated retinal pigment epithelium (RPE) segmentation was performed, followed by manual grading of drusen by two independent experts.
• Segmentation errors were classified, and correlations between automated and manual drusen identification were analyzed based on druse dimensions.

Main Results:

• A total of 1356 drusen were analyzed.
• The Cirrus device demonstrated significantly fewer segmentation errors compared to the 3DOCT-1000 (P < 0.001).
• The Cirrus 200 × 200 scan pattern detected 30% of drusen with negligible errors; Spectralis lacked true RPE segmentation. Expert grading yielded higher drusen counts than fundus photographs (P < 0.05).

Conclusions:

• SD-OCT imaging effectively visualizes drusen-related RPE disease in AMD.
• Current automated segmentation algorithms exhibit limitations in reliably identifying the quantity and actual size of drusen, especially smaller ones.
• Further refinement of automated algorithms is necessary for precise drusen quantification in clinical practice.