Low-Dose CT Denoising Using a Structure-Preserving Kernel Prediction Network

TL;DR

This paper introduces StructKPN, a novel deep learning model for low-dose CT denoising that preserves detailed structures by using a structure-aware loss and spatially-variant filtering, outperforming existing methods.

Contribution

The paper presents a structure-preserving kernel prediction network with a structure-aware loss, improving noise removal while maintaining fine details in low-dose CT images.

Findings

Achieved superior denoising performance on synthetic and real datasets.

Better preservation of fine structures compared to traditional CNN approaches.

Effective in clinical screening and low-dose protocol optimization.

Abstract

Low-dose CT has been a key diagnostic imaging modality to reduce the potential risk of radiation overdose to patient health. Despite recent advances, CNN-based approaches typically apply filters in a spatially invariant way and adopt similar pixel-level losses, which treat all regions of the CT image equally and can be inefficient when fine-grained structures coexist with non-uniformly distributed noises. To address this issue, we propose a Structure-preserving Kernel Prediction Network (StructKPN) that combines the kernel prediction network with a structure-aware loss function that utilizes the pixel gradient statistics and guides the model towards spatially-variant filters that enhance noise removal, prevent over-smoothing and preserve detailed structures for different regions in CT imaging. Extensive experiments demonstrated that our approach achieved superior performance on both synthetic and non-synthetic datasets, and better preserves structures that are highly desired in clinical screening and low-dose protocol optimization.