Development of Automated Methods For Retinal Optical Coherence Tomography Image Analysis

Abstract

Cystoid Macular Edema (CME) is a retinal abnormality causing fluid accumulations in the retina due to various retinal diseases. The early diagnosis of CME and its quan- tification is vital for treatment planning. Optical Coherence Tomography (OCT) is a non-invasive imaging technique used to visualize the human retina and the retinal ab- normalities. The OCT images have corrupted with speckle noise due to the coherence detection, which will degrade the quality of the OCT images. Also, the human retina is a layered structure. The segmentation of retinal layers helps in diagnose various retinal diseases and finding the locations of retinal pathologies. This thesis focused on the de- velopment of automated methods for retinal OCT image analysis and tried to provide deep learning-based solutions for each of the stages. OCT is an imaging technique widely used for medical imaging. Noise in an OCT image generally degrades its quality, thereby obscuring clinical features and making the automated segmentation task suboptimal. Obtaining higher quality images requires sophisticated equipment and technology, available only in selected research settings, and is expensive to acquire. Developing effective denoising methods to improve the quality of the images acquired on systems currently in use has the potential for vastly improving image quality and automated quantitative analysis. Noise characteristics in images acquired from machines of different makes and models may vary. Our experi- ments show that any single state-of-the-art method for noise reduction fails to perform equally well on images from various sources. Therefore, detailed analysis is required to determine the exact noise type in images acquired using different OCT machines. In the second chapter, we studied noise characteristics in the publicly available DUKE and OPTIMA datasets to build a more efficient model for noise reduction. These datasets have OCT images acquired using machines of different manufacturers. We further pro- iii pose a patch-wise training methodology to build a system to effectively denoise OCT images. We have performed an extensive range of experiments to show that the pro- posed method performs superior to other state-of-the-art methods. Segmentation of retinal layers is a vital step in computerized processing and the study of retinal OCT images. However, automatic segmentation of retinal layers is chal- lenging due to the presence of noise, widely varying reflectivity of image components, variations in morphology and alignment of layers in the presence of retinal diseases. In the third chapter, we propose a Fully Convolutional Network (FCN) termed as DelNet based on a deep ensemble learning approach to selectively segment retinal layers from OCT scans. The proposed model is tested on a publicly available DUKE DME dataset. Comparative analysis with other state-of-the-art methods on a benchmark dataset shows that the performance of DelNet is superior to other methods. In the fourth chapter, we propose attention assisted convolutional neural network- based architecture to detect and quantify three types of retinal cysts namely the intra- retinal cyst, sub-retinal cyst and pigmented epithelial detachment from the OCT images of the human retina. The proposed architecture has an encoder-decoder structure with an attention and a multi-scale module. The qualitative and quantitative performance of the model is evaluated on the publicly available RETOUCH retinal OCT fluid detection challenge data set. The proposed model outperforms the state-of-the-art methods in terms of precision, recall, and dice coefficient. Furthermore, the proposed model is computationally efficient due to its less number of model parameters.