Statistical CT reconstruction using region-aware texture preserving regularization learning from prior normal-dose CT...

Area of Science:

• Medical Imaging
• Image Reconstruction
• Computational Imaging

Background:

• Prior normal-dose CT (NdCT) images contain valuable texture and structure information.
• This information can potentially enhance follow-up low-dose CT (LdCT) image reconstruction.
• Current reconstruction methods may not fully leverage this prior information.

Purpose of the Study:

• To develop a novel LdCT image reconstruction method that learns and utilizes texture information from available NdCT images.
• To preserve essential textures and structure features in LdCT images.
• To improve the diagnostic quality of LdCT scans.

Main Methods:

• A region-aware texture preserving (RATP) prior was developed.
• The method learns texture information from similar patches identified in NdCT images.
• Similar patches are clustered using efficient context feature searching.
• Redundant texture information from similar patches serves as a priori knowledge for LdCT reconstruction.

Main Results:

• The RATP method demonstrated superior performance in LdCT imaging.
• Qualitative and quantitative evaluations were performed using patient data.
• RATP outperformed filtered back projection (FBP) and statistical iterative reconstruction (SIR) methods with various regularization techniques.

Conclusions:

• The proposed RATP prior effectively learns texture features from NdCT images.
• RATP adaptively characterizes region-specific structures in LdCT images.
• The method significantly enhances LdCT image quality and detail preservation compared to existing techniques.