Head-Neck Dual-energy CT Contrast Media Reduction Using Diffusion Models

TL;DR

This paper introduces a diffusion model-based deep learning approach to reduce contrast media in dual-energy CT scans, achieving high-quality images with significantly lower contrast agent doses, addressing supply shortages and safety concerns.

Contribution

The study presents a novel diffusion model framework for contrast media reduction in DECT, outperforming existing methods and enabling effective imaging with minimal contrast agent use.

Findings

High-quality images generated with only 12.5% contrast dose

Outperforms competing methods in human reader study

Addresses contrast media shortage and safety issues

Abstract

Iodinated contrast media is essential for dual-energy computed tomography (DECT) angiography. Previous studies show that iodinated contrast media may cause side effects, and the interruption of the supply chain in 2022 led to a severe contrast media shortage in the US. Both factors justify the necessity of contrast media reduction in relevant clinical applications. In this study, we propose a diffusion model-based deep learning framework to address this challenge. First, we simulate different levels of low contrast dosage DECT scans from the standard normal contrast dosage DECT scans using material decomposition. Conditional denoising diffusion probabilistic models are then trained to enhance the contrast media and create contrast-enhanced images. Our results demonstrate that the proposed methods can generate high-quality contrast-enhanced results even for images obtained with as low as 12.5% of the normal contrast dosage. Furthermore, our method outperforms selected competing methods in a human reader study.