Image Enhancement of Optical Coherence Tomography using Deep Learning

Optical coherence tomography (OCT) images are widely used in the clinical diagnosis of diseases because they can obtain high-resolution images in real-time. However, due to the noise interference generated during the signal acquisition, there will be pixel jitters in the OCT images. Aiming at the pixel imaging jitter problem caused by signal transmission interference, an improved UNET network framework in deep learning is proposed to construct an OCT image correction model. This model forms a mapping from the input image X to the output image Y by taking advantage of the deep network structure of UNET. Through 200 iteration training, the loss value is reduced to the lowest level in this model to realize OCT image correction. Finally, the validity of the proposed method was proved by calculating the similarity of corrected images.