Inter-expert reliability in multi-field-of-view automatic drusen segmentation analysis using optical coherence tomography

Abstract

[Abstract]: Drusen are commonly associated with age-related macular degeneration (AMD), a leading cause of blindness in older adults. They can vary significantly in size, appearance, and location within the retina, impacting macular health in various ways. Smaller drusen may not affect the vision significantly, but larger and more numerous drusen can determine more severe AMD and a higher risk of progressing to late-stage disease, which can lead to significant visual impairment. Optical Coherence Tomography (OCT) is a critical imaging technique in the field of ophthalmology, particularly in the study and management of retinal diseases. The use of different fields of view (FoVs) in OCT imaging plays a pivotal role in enhancing our understanding and management of various retinal conditions, especially drusen. To address a crucial gap in the literature, this study introduces a novel fully-automatic approach for segmenting drusen in OCT images, applying FoV analysis for the first time to assist clinicians in diagnosing ocular diseases. To achieve this, we analyzed three different datasets, utilizing deep learning to enable a comprehensive comparison of segmentation accuracy across various FoVs and introducing a new method to assess inter-expert agreement in this challenging domain. Additionally, our research pioneers the segmentation of the 3×3mm area, essential for identifying significant retinal changes. Our approach not only closely aligns with expert assessments but also represents a significant step towards standardizing diagnostic procedures in ophthalmology, enhancing both the precision and efficiency of retinal image analysis to aid clinical decision-making.