Cross-Domain Denoising for Low-Dose Multi-Frame Spiral Computed Tomography

Area of Science:

• Medical Imaging
• Radiology
• Computer Vision

Background:

• Computed tomography (CT) is a vital diagnostic tool, but ionizing radiation poses health risks.
• Reducing radiation dose in CT (low-dose CT or LDCT) is crucial for patient safety.
• Existing LDCT denoising methods often perform poorly on real-world data due to domain differences from simulations.

Purpose of the Study:

• To develop an effective LDCT denoising method for multi-slice spiral CT scanners.
• To address the limitations of existing methods, particularly their performance on real-world data.
• To improve image quality and diagnostic confidence in low-dose CT scans.

Main Methods:

• A novel two-stage denoising method is proposed, specifically designed for commercially available multi-slice spiral CT scanners.
• The method leverages the redundancy in multi-slice projections and volumetric data.
• It addresses the over-smoothing issue common in conventional cascaded denoising frameworks.

Main Results:

• The proposed method achieved significant noise reduction (up to 70%) in LDCT images.
• Crucially, the method preserved spatial resolution, avoiding the common trade-off with noise reduction.
• Radiologist evaluations confirmed superior performance compared to state-of-the-art methods in clinical practice.

Conclusions:

• The developed two-stage method offers a robust solution for LDCT denoising in real-world clinical applications.
• It effectively balances noise reduction and preservation of essential image details.
• This approach has the potential to enhance diagnostic accuracy while minimizing radiation exposure.