Degradation-Aware Deep Learning Framework for Sparse-View CT Reconstruction

Area of Science:

• Medical Imaging
• Computer Vision
• Deep Learning

Background:

• Sparse-view CT reconstruction is crucial for reducing artifacts and enhancing image details in degraded CT scans.
• Current deep learning methods struggle with inconsistent degradation levels between training and testing data.
• Existing approaches require extensive storage for multiple models or are ineffective with varied degradation.

Purpose of the Study:

• To develop a single, scalable deep learning framework for sparse-view CT reconstruction across multiple degradation levels.
• To address the limitations of existing methods in handling varied image degradation strengths.
• To improve the effectiveness and efficiency of CT image reconstruction.

Main Methods:

• A novel degradation-aware deep learning framework is proposed.
• The framework utilizes a dual-domain approach, analyzing degradation in both frequency and image domains.
• Specific operations are applied at different degradation levels for frequency component recovery and spatial detail reconstruction.

Main Results:

• The proposed method demonstrates superior performance compared to classical deep learning reconstruction techniques.
• Quantitative metrics such as Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity (SSIM) show significant improvements.
• Visual results confirm enhanced effectiveness and scalability in reconstructing clear CT images.

Conclusions:

• The degradation-aware deep learning framework offers an effective and scalable solution for sparse-view CT reconstruction.
• The dual-domain approach successfully addresses challenges posed by varying degradation levels.
• This method advances the field by providing a unified solution for diverse CT image quality scenarios.