Development of novel algorithms suitable for clinical assessment of ocular diseases using OCT imaging

Abstract

Optical coherence tomography (OCT) is a widely utilized imaging modality in ophthalmology. However, its accuracy and clinical utility are limited by several significant challenges. This study describes a series of innovative signal processing approaches and novel algorithms for OCT and OCT angiography (OCTA) that address these limitations and enhance the accuracy and comprehensiveness of imaging for the diagnosis and monitoring of ocular diseases.Our first contribution is a three-dimensional registration algorithm that enhances the contrast and signal-to-noise ratio of OCTA images, thereby facilitating improved visualization and quantification of the choriocapillaris. Additionally, we propose a novel model for widefield OCTA distortion correction and demonstrate that this correction significantly improves quantification metrics. Subsequently, we differentiate between various types of drusen and their interaction of with the choriocapillaris, and analyze the grading exercise of macular atrophy. This study demonstrated how OCT can be utilized to identify and monitor the biomarkers of age-related macular degeneration accurately. To further advance AMD diagnosis and monitoring, a deep learning model for geographic atrophy (GA) segmentation based on hyper-transmission defects is presented. This is complemented by a method for automatic identification, segmentation, and comparison of hyperpigmentation extent using optical attenuation coefficients (OAC) obtained from both swept-source (SS) OCT and spectral domain (SD) OCT scans. By enhancing our understanding of these biomarkers, OCT and OCTA can act a critical role in quantifying the progression of this disease. Together, these methods represent a significant advancement in OCT image analysis, holding immense promise for improved clinical diagnosis and monitoring of ocular diseases.