Automated Three-Dimensional Image Segmentation of Retinal OCT Images

Optical coherence tomography (OCT) has become an essential tool in\ud ophthalmology. The macular OCT images are three dimensional, high resolution and\ud cross-sectional image of macula, which provides detailed structural information of\ud various macular layers. Moreover macular OCT imaging is non-contact, noninvasive\ud and real-time. OCT images can be obtained quickly and multiple times, this\ud allows easy diagnosis and treatment progress monitoring. With advances in\ud technology acquired dataset size is also increasing, which makes it impractical to\ud manually segment macular layers in clinical environment. Thus need of fast and\ud accurate segmentation algorithm is ever increasing. In this thesis two segmentation\ud algorithms have been developed. The two-dimensional segmentation algorithm\ud based on gradient method successfully marked three retinal layers and five layer\ud boundaries. The three-dimensional segmentation algorithm is based on threshold and\ud characteristic intensity profile of macular OCT scan. The gradient method is well\ud suited for boundary detection where large changes in intensity occur