Noise as Domain Shift: Denoising Medical Images by Unpaired Image Translation

TL;DR

This paper presents a novel unpaired image translation approach using a modified cycleGAN to effectively denoise retinal OCT images by modeling the problem as domain translation between high and low noise images.

Contribution

It introduces a cycleGAN-based method for unpaired denoising of medical images, outperforming existing techniques and learning to distinguish subtle noise levels and retinal structures.

Findings

Outperforms established denoising methods in quantitative and qualitative evaluations.

Learns to differentiate subtle noise variations and retinal layers.

Effectively models noise as a domain shift in medical images.

Abstract

We cast the problem of image denoising as a domain translation problem between high and low noise domains. By modifying the cycleGAN model, we are able to learn a mapping between these domains on unpaired retinal optical coherence tomography images. In quantitative measurements and a qualitative evaluation by ophthalmologists, we show how this approach outperforms other established methods. The results indicate that the network differentiates subtle changes in the level of noise in the image. Further investigation of the model's feature maps reveals that it has learned to distinguish retinal layers and other distinct regions of the images.