Knowledge Bank

Glaucoma is a leading cause of irreversible blindness, affecting over 80 million individuals globally. Intraocular pressure (IOP) is a primary risk factor, but how IOP affects the morphometry of the posterior eye where glaucoma damages occur is not well understood. This research focuses on two primary aims: (1) investigating the relationship between IOP and lamina cribrosa (LC) morphology, and (2) improving ocular ultrasound image quality using deep learning-based denoising algorithms to better extract morphological features. Various deep learning architectures, including convolutional neural networks, autoencoders, and diffusion models were evaluated for denoising ultrasound images. Performance was assessed using metrics such as peak signal-to-noise ratio (PSNR) and contrast-to-noise ratio (CNR). Ultrasound images of the posterior eye were previously acquired from human donor eyes controlling IOP from 5 to 30 mmHg. LC thickness and anterior lamina cribrosa surface depth (ALCSD) changes were analyzed in manually segmented ultrasound images. Results indicate that LC thickness decreased with increasing IOP, while ALCSD did not significantly change. Denoising efforts were made using deep learning architectures like autoencoders and diffusion models, but did not result in any improvement. A fine-tuned vision transformer was able to accurately generate LC masks, enabling automated segmentation of ocular ultrasound. This work contributes to the understanding of pressure-related morphological changes in the posterior eye and lays a foundation for more robust detection of posterior eye changes associated with IOP elevation.