A benchmark study of automated intra-retinal cyst segmentation algorithms using optical coherence tomography B-scans

Area of Science:

• Ophthalmic image analysis
• Medical image segmentation
• Retinal imaging technology

Background:

• Retinal cysts, fluid accumulation in the retina, are key indicators of ocular diseases like macular degeneration and diabetic macular edema.
• Accurate segmentation and quantification of intra-retinal cysts are vital for diagnosing retinal pathology and assessing disease severity.
• Automated segmentation of intra-retinal cysts from optical coherence tomography (OCT) B-scans is increasingly important in retinal image analysis.

Purpose of the Study:

• To compare and benchmark various automated intra-retinal cyst segmentation algorithms.
• To analyze the performance and scalability of these algorithms across different image acquisition systems.

Main Methods:

• A modular approach was used to standardize diverse segmentation algorithms.
• Performance variations were analyzed using the OPTIMA cyst segmentation challenge dataset.
• Quantitative and qualitative experiments were conducted to compare key automated methods.

Main Results:

• Significant variations in automated cyst segmentation performance were observed.
• Factors such as signal-to-noise ratio (SNR), retinal layer morphology, and post-processing steps critically influence segmentation accuracy.
• Algorithm scalability across different image acquisition systems was analyzed.

Conclusions:

• Benchmarking provides insights into the scalability of automated segmentation across vendor-specific imaging modalities.
• This study offers guidance for improving retinal pathology diagnostics and treatment processes through automated analysis.
• Understanding the impact of image quality and processing steps is essential for reliable automated cyst segmentation.