Sparse-View CT Joint Reconstruction Strategy with Sparse Sampling Encoding Layer

Area of Science:

• Medical Imaging
• Computer Vision
• Artificial Intelligence

Background:

• Sparse angular projection is crucial for reducing CT radiation dose.
• Traditional reconstruction methods struggle with sparse projection angles, leading to image degradation.
• Existing deep learning methods improve reconstruction but require predefined sampling schemes.

Purpose of the Study:

• To develop an automated method for selecting efficient sparse sampling schemes under dose constraints.
• To create an end-to-end sparse angular CT reconstruction approach.

Main Methods:

• A novel sampling encoding layer was developed to search for optimal sparse sampling schemes.
• This layer was integrated into a sparse reconstruction neural network.
• A joint reconstruction strategy utilizing both radon and image domains was employed.

Main Results:

• Experiments on public CT datasets confirmed the method's effectiveness.
• The proposed approach demonstrated successful sparse angular CT reconstruction.

Conclusions:

• The developed joint reconstruction network with a sparse sampling coding layer shows significant potential for clinical application.
• This method offers a promising solution for dose reduction in CT imaging.