3D Segmentation & Measurement of Macular Holes

Abstract

Macular holes are blinding conditions where a hole develops in the central part of retina, resulting in reduced central vision. The prognosis and treatment options are related to a number of variables including the macular hole size and shape. In this work we introduce a method to segment and measure macular holes in three-dimensional (3D) data.

High-resolution spectral domain optical coherence tomography (SD-OCT) allows precise imaging of the macular hole geometry in three dimensions, but the measurement of these by human observers is time consuming and prone to high inter- and intra-observer variability, being characteristically measured in 2D rather than 3D. This work introduces several novel techniques to automatically retrieve accurate 3D measurements of the macular hole, including surface area, base area, base diameter, top area, top diameter, height, and minimum diameter. Specifically, it is introducing a multi-scale 3D level set segmentation approach based on a state-of-the-art level set method, and introducing novel curvature-based cutting and 3D measurement procedures. The algorithm is fully automatic, and we validate the extracted measurements both qualitatively and quantitatively, where the results show the method to be robust across a variety of scenarios.

A segmentation software package is presented for targeting medical and biological applications, with a high level of visual feedback and several usability enhancements over existing packages. Specifically, it is providing a substantially faster graphics processing unit (GPU) implementation of the local Gaussian distribution fitting (LGDF) energy model, which can segment inhomogeneous objects with poorly defined boundaries as often encountered in biomedical images. It also provides interactive brushes to guide the segmentation process in a semi-automated framework. The speed of implementation allows us to visualise the active surface in real-time with a built-in ray tracer, where users may halt evolution at any timestep to correct implausible segmentation by painting new blocking regions or new seeds. Quantitative and qualitative validation is presented, demonstrating the practical efficacy of the interactive elements for a variety of real-world datasets.

The size of macular holes is known to be one of the strongest predictors of surgical success both anatomically and functionally. Furthermore, it is used to guide the choice of treatment, the optimum surgical approach and to predict outcome. Our automated 3D image segmentation algorithm has extracted 3D shape-based macular hole measurements and described the dimensions and morphology. Our approach is able to robustly and accurately measure macular hole dimensions.

This thesis is considered as a significant contribution for clinical applications particularly in the field of macular hole segmentation and shape analysis.