Analysis of optical coherence tomography images using deep convolutional neural network for maculopathy grading

Blindness is the inability to see and it is majorly caused due to retinal diseases. There are many eye testing techniques which are used to detect retinal disorders, but optical coherence tomography (OCT) is the most widely used technique due to its ability to detect retinal abnormalities in early stages. Many researchers have worked on extracting retinal information from OCT images. However, to the best of our knowledge, there is no literature available that presents a robust architecture for the automated diagnosis and clinical grading of maculopathy irrespective of the acquisition machinery. Therefore, this chapter presents a deep convolutional neural network architecture (TU-Net) for the grading of maculopathy. The proposed TU-Net architecture has been trained on 4992 retinal OCT scans and tested on 41,549 retinal OCT scans acquired from Topcon 3D OCT 2000 and Spectralis Heidelberg OCT machines. These scans are taken from publicly available Duke datasets and the Biomedical Image and Signal Analysis (BIOMISA) dataset where TU-Net achieved the accuracy of 0.9342 for grading maculopathy as per clinical standards.