Contextual Checkerboard Denoise -- A Novel Neural Network-Based Approach for Classification-Aware OCT Image Denoising

TL;DR

This paper tackles the challenge of denoising medical OCT images without compromising diagnostically relevant features, addressing the impracticality of obtaining ground-truth clean images for supervised training. It introduces Contextual Checkerboard Denoise, a self-supervised, classification-aware framework built on a ResUNet++ backbone, employing checkerboard blind-spotting to learn from noisy data while jointly optimizing denoising and classification loss. A dual-model training scheme predicts odd and even pixels, with fusion at inference, and a mutual learning setup between CNN and ViT classifiers further enhances classification performance on denoised outputs. Empirical results on the VIP Cup 2024 OCT dataset show superior denoising metrics and classification accuracy compared with baselines, demonstrating practical impact for improved OCT-based diagnosis and broader applicability to medical imaging denoising tasks.

Abstract

In contrast to non-medical image denoising, where enhancing image clarity is the primary goal, medical image denoising warrants preservation of crucial features without introduction of new artifacts. However, many denoising methods that improve the clarity of the image, inadvertently alter critical information of the denoised images, potentially compromising classification performance and diagnostic quality. Additionally, supervised denoising methods are not very practical in medical image domain, since a \emph{ground truth} denoised version of a noisy medical image is often extremely challenging to obtain. In this paper, we tackle both of these problems by introducing a novel neural network based method -- \emph{Contextual Checkerboard Denoising}, that can learn denoising from only a dataset of noisy images, while preserving crucial anatomical details necessary for image classification/analysis. We perform our experimentation on real Optical Coherence Tomography (OCT) images, and empirically demonstrate that our proposed method significantly improves image quality, providing clearer and more detailed OCT images, while enhancing diagnostic accuracy.

Figures (4)

• Figure 1: Checkerboard model architecture
• Figure 2: Checkerboard blind-spot creation
• Figure 3: Mutual learning network for classification
• Figure 4: Visual comparison of denoising results among existing models and our proposed model.