Agreement between Routine-Dose and Lower-Dose CT with and without Deep Learning-based Denoising for Active...

Area of Science:

• Radiology
• Medical Imaging
• Oncology

Background:

• Active surveillance is crucial for managing small renal masses (SRMs).
• Evaluating SRMs requires high-quality contrast-enhanced CT scans.
• Reducing radiation dose is a key goal in medical imaging.

Purpose of the Study:

• To assess the agreement between routine-dose (RD) and lower-dose (LD) contrast-enhanced CT scans for evaluating SRMs.
• To evaluate the impact of deep learning-based denoising (DLD) on LD CT image quality and diagnostic accuracy.
• To determine the feasibility of dose reduction in CT for SRMs during active surveillance.

Main Methods:

• Retrospective analysis of CT scans from patients with SRMs under active surveillance.
• Generation of LD CT images simulating 75% (LD75) and 90% (LD90) dose reductions from RD images.
• Application of DLD to LD images (LD75-DLD, LD90-DLD).
• Independent evaluation of 350 CT scans by nine radiologists across five datasets for tumor size, collecting system proximity, and shape irregularity.
• Assessment of interobserver reproducibility and agreement using LOAM and Gwet AC2 coefficients.
• Subjective and quantitative image quality assessments.

Main Results:

• LD75 CT showed agreement with RD scans for SRM diameter assessment (LOAM ±2.4 mm vs. ±2.2 mm).
• A 90% dose reduction compromised reproducibility (LOAM ±3.0 mm), but LD90-DLD preserved it (LOAM ±2.4 mm).
• Observer agreement for tumor nearness to collecting system and shape irregularity was comparable across all image sets (P ≥ .35 for TN, P ≥ .02 for TSI).
• DLD effectively restored image quality at reduced doses; LD75-DLD demonstrated superior image quality, lower noise, and improved contrast-to-noise ratio compared to RD (P < .001).

Conclusions:

• A 75% radiation dose reduction is feasible for SRM assessment in active surveillance using CT with conventional iterative reconstruction.
• Deep learning-based denoising (DLD) enables sub-millisievert dose CT scans for evaluating SRMs.
• DLD effectively maintains diagnostic performance and image quality at significantly reduced radiation doses.