Exploring the Low-Dose Limit for Focal Hepatic Lesion Detection with a Deep Learning-Based CT Reconstruction...

Area of Science:

• Medical Imaging
• Radiology
• Artificial Intelligence

Background:

• Computed tomography (CT) is crucial for diagnosing hepatic lesions.
• Radiation dose reduction in CT is an ongoing challenge.
• Deep learning algorithms offer potential for improving low-dose CT imaging.

Purpose of the Study:

• To evaluate the dose reduction capabilities of artificial intelligence iterative reconstruction (AIIR) for hepatic lesion detection.
• To compare the performance of AIIR against hybrid iterative reconstruction (HIR) at reduced radiation doses.
• To assess the impact of AIIR on lesion conspicuity, diagnostic confidence, and image quality.

Main Methods:

• Retrospective analysis of 40 patients with 98 hepatic lesions.
• Simulated low-dose CT images at 40%, 20%, and 10% of routine dose.
• Reconstruction using both AIIR and HIR algorithms.
• Qualitative assessment by two radiologists and quantitative analysis of contrast-to-noise ratio (CNR).

Main Results:

• AIIR demonstrated significantly higher lesion CNR compared to routine-dose HIR at reduced doses (p < 0.05).
• No significant difference in image quality between 40%-dose AIIR and routine-dose HIR.
• AIIR achieved 100% detection rate at 40% dose, outperforming HIR (98%).
• AIIR lesion detection rates were superior to HIR across all simulated low-dose levels.

Conclusions:

• AIIR enables substantial radiation dose reduction (up to 60%) in liver CT while preserving diagnostic image quality.
• AIIR significantly enhances lesion detection performance in low-dose CT compared to conventional HIR.
• AIIR represents a promising advancement for safer and more effective hepatic lesion detection using CT.