Disentangled deep learning method for interior tomographic reconstruction of low-dose x-ray CT.

Area of Science:

• Medical Imaging
• Computational Imaging
• Artificial Intelligence in Medicine

Background:

• Low-dose interior tomography combines low-dose CT (LDCT) with region-of-interest (ROI) imaging for dose reduction and high-resolution imaging.
• Challenges in accurate tomographic reconstruction arise from the combined effects of noise and data truncation.
• Developing novel reconstruction frameworks is crucial for high-quality ROI imaging and efficient extension of recoverable regions.

Purpose of the Study:

• To develop a novel deep learning-based reconstruction framework for low-dose interior tomography.
• To address the coupled ill-posed problems caused by noise and truncated projections.
• To achieve high-quality ROI reconstruction and efficient extension of the recoverable region.

Main Methods:

• A comprehensive analysis of projection data composition and angular sampling patterns in low-dose interior tomography was conducted.
• Two deep learning-based reconstruction pipelines were proposed: Deep Projection Extraction-based Reconstruction (DPER) and DPER with Progressive extension (DPER-Pro).
• DPER utilizes a dual-domain deep neural network to disentangle and extract noise and background projection contributions for ROI reconstruction. DPER-Pro enhances DPER with a progressive "coarse-to-fine" strategy for missing data compensation.

Main Results:

• DPER effectively handles coupled ill-posed problems, achieving high-quality ROI reconstructions by accurately extracting noise and background projections.
• DPER-Pro extends the recoverable region while preserving ROI image quality by leveraging disentangled projection components and angular sampling patterns.
• Both methods outperform competing approaches in reconstructing reliable structures, enhancing generalization, and mitigating noise and truncation artifacts.

Conclusions:

• The proposed decoupled deep learning framework offers a robust solution for low-dose interior tomography, effectively addressing challenges from noise and truncated projections.
• The methods significantly improve ROI reconstruction quality and efficiently recover structural information in exterior regions.
• This work presents a promising pathway for advancing low-dose ROI imaging in various applications.