Reduced-dose dual-energy CT with deep learning image reconstruction for detection and characterization of liver metastases

Abstract

Purpose

To compare image quality and diagnostic performance of reduced-dose dual-energy CT (DECT) with deep learning image reconstruction (DLIR) versus standard-dose single-energy CT (SECT) with adaptive statistical iterative reconstruction-Veo (ASIR-V) for detecting liver metastases, and to evaluate the efficacy of differentiating metastases from cysts using DECT spectral parameters.

Methods

Eighty participants with known or suspected liver metastases from June 2023 to January 2025 were prospectively enrolled and underwent contrast-enhanced liver CT with either standard-dose SECT (n = 40, 120-kVp images, ASIR‑V 40 %) or reduced-dose DECT (n = 40, 40- and 70-keV virtual monoenergetic images [VMIs], high-intensity DLIR [DH]). The objective image noise, contrast-to-noise ratio (CNR), signal-to-noise ratio (SNR), liver-to-lesion contrast-to-noise ratio (LLR), subjective image quality, lesion conspicuity, and detection rate were assessed. The diagnostic performance of spectral parameters for differentiating metastases from cysts was evaluated using receiver operating characteristic (ROC) curves.

Results

DH significantly reduced image noise of DECT scans in reduced radiation dose conditions. With a 45 % dose reduction, the 40- and 70-keV VMIs with DH showed higher CNR, SNR, and LLR, better image quality and similar lesion detection rates and better or comparable lesion conspicuity compared with AR40 120-kVp images (P > 0.05). The use of the spectral curve slope, iodine concentration, normalized iodine concentration, and effective atomic number yielded the area under the curve (AUC) values of 0.977, 0.990, 0.982, and 0.980 for differentiating metastases from cysts, respectively.

Conclusion

DLIR effectively reduces image noise and improves image quality of the 40- and 70-keV VMIs in DECT, achieving a 45% radiation dose reduction without compromising metastases diagnosis. DECT spectral parameters enable accurate differentiation of metastases from cysts.