Adversarial training with cycle consistency for unsupervised super-resolution in endomicroscopy

TL;DR

This paper introduces an unsupervised adversarial cycle consistency framework for super-resolution in endomicroscopy images, overcoming the lack of paired high- and low-resolution data to enhance image quality for medical diagnostics.

Contribution

The proposed method is the first to apply unsupervised adversarial training with cycle consistency for super-resolution in endomicroscopy, enabling quality transfer from high-resolution images without paired data.

Findings

Produced convincing super-resolved images validated quantitatively

Achieved high Mean Opinion Score in subjective quality assessment

Effective in real clinical endomicroscopy datasets

Abstract

In recent years, endomicroscopy has become increasingly used for diagnostic purposes and interventional guidance. It can provide intraoperative aids for real-time tissue characterization and can help to perform visual investigations aimed for example to discover epithelial cancers. Due to physical constraints on the acquisition process, endomicroscopy images, still today have a low number of informative pixels which hampers their quality. Post-processing techniques, such as Super-Resolution (SR), are a potential solution to increase the quality of these images. SR techniques are often supervised, requiring aligned pairs of low-resolution (LR) and high-resolution (HR) images patches to train a model. However, in our domain, the lack of HR images hinders the collection of such pairs and makes supervised training unsuitable. For this reason, we propose an unsupervised SR framework based on an adversarial deep neural network with a physically-inspired cycle consistency, designed to impose some acquisition properties on the super-resolved images. Our framework can exploit HR images, regardless of the domain where they are coming from, to transfer the quality of the HR images to the initial LR images. This property can be particularly useful in all situations where pairs of LR/HR are not available during the training. Our quantitative analysis, validated using a database of 238 endomicroscopy video sequences from 143 patients, shows the ability of the pipeline to produce convincing super-resolved images. A Mean Opinion Score (MOS) study also confirms this quantitative image quality assessment.

Figures