Ultra-low-dose CT for malignant metastasis screening using a deep learning image reconstruction algorithm.

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Cancer surveillance requires accurate imaging to detect metastatic disease.
• Ultra-low-dose CT (ULD-CT) aims to reduce radiation exposure during surveillance.
• Deep learning reconstruction techniques offer potential for improving ULD-CT image quality.

Purpose of the Study:

• To evaluate the feasibility and diagnostic performance of super-resolution deep learning reconstruction (SR-DLR) for ULD-CT in cancer surveillance.
• To compare SR-DLR with normal-resolution DLR (NR-DLR) and hybrid iterative reconstruction (HIR).

Main Methods:

• Prospective enrollment of cancer patients undergoing surveillance.
• Contrast-enhanced whole-body ULD-CT performed with images reconstructed using HIR, NR-DLR, and SR-DLR.
• Independent evaluation of image quality, lesion detectability, and diagnostic performance by two radiologists.

Main Results:

• ULD-CT with SR-DLR achieved approximately 70% radiation dose reduction compared to standard-dose CT.
• SR-DLR significantly improved image quality, sharpness, and noise reduction versus HIR and NR-DLR (p < 0.001).
• SR-DLR demonstrated superior detection rates for malignant and benign lesions, maintaining diagnostic accuracy.

Conclusions:

• SR-DLR is a feasible technique for ULD-CT in oncological patients.
• SR-DLR enables substantial radiation dose reduction while preserving/enhancing image quality and diagnostic performance.
• This method is effective for detecting visceral and soft-tissue metastatic disease during cancer surveillance.