Early Detection of Macular Edema using 3D OCT Images and Deep Learning Algorithms
Abstract
The occurrence of Macular Edema, which involves the accumulation of fluids within the Macula, could result in the loss of your eyesight if it is not diagnosed quickly enough. Optical Coherence Tomography (OCT) has transformed retinal imaging into 3D images that provide detail about the layers of the retina. Using Three Dimensional (3-Dimensional) OCT Imaging adds another level of detail by producing volumetric data; therefore, the retinal abnormality could be assessed more completely than before. In this research, we develop a new deep learning framework using 3-Dimensional Convolutions Neural Networks (3D-CNNs) for identifying Macular Edema at an early stage from 3-Dimensional OCT Images. We show that our model is able to capture spatial details present in volumetric data, thus improving the diagnostic precision. After evaluating rigorously, the results demonstrated better Sensitivity and Specificity when comparing to both traditional 2D-CNN and Machine Learning methods. The results demonstrate that there may be possible integration of Deep Learning Methods and 3-Dimensional OCT Images for facilitating earlier diagnosis of patients, improvement of patient outcomes, and the streamlining of Clinical Workflows.