Patch-wise Deep Metric Learning for Unsupervised Low-Dose CT Denoising

TL;DR

This paper introduces an unsupervised patch-wise deep metric learning method for low-dose CT denoising that preserves CT number accuracy while effectively reducing noise.

Contribution

It proposes a novel unsupervised learning framework that maintains CT number fidelity using patch-wise deep metric learning, addressing limitations of existing methods.

Findings

Produces high-quality denoised images without CT number shift

Outperforms traditional unsupervised denoising methods

Effectively suppresses noise while preserving anatomical structures

Abstract

The acquisition conditions for low-dose and high-dose CT images are usually different, so that the shifts in the CT numbers often occur. Accordingly, unsupervised deep learning-based approaches, which learn the target image distribution, often introduce CT number distortions and result in detrimental effects in diagnostic performance. To address this, here we propose a novel unsupervised learning approach for lowdose CT reconstruction using patch-wise deep metric learning. The key idea is to learn embedding space by pulling the positive pairs of image patches which shares the same anatomical structure, and pushing the negative pairs which have same noise level each other. Thereby, the network is trained to suppress the noise level, while retaining the original global CT number distributions even after the image translation. Experimental results confirm that our deep metric learning plays a critical role in producing high quality denoised images without CT number shift.