Edge-oriented dual-dictionary guided enrichment (EDGE) for MRI-CT image reconstruction

Area of Science:

• Medical Imaging
• Computational Imaging
• Biomedical Engineering

Background:

• Accurate Magnetic Resonance Imaging (MRI) reconstruction is crucial for diagnosis.
• Reconstruction from under-sampled data presents significant challenges.
• Integrating data from different imaging modalities like Computed Tomography (CT) offers potential solutions.

Purpose of the Study:

• To develop a novel algorithm for joint/simultaneous X-ray CT and MRI image reconstruction.
• To improve MRI image reconstruction from highly under-sampled k-space data using CT images.
• To establish a robust correspondence between CT and MRI modalities.

Main Methods:

• A two-step algorithm involving dictionary learning (DL) and edge-oriented dual-dictionary guided enrichment (EDGE).
• Generation of a training dataset from registered CT-MRI pairs.
• Reconstruction of an initial MRI estimate using EDGE and CT data, followed by DL-based reconstruction from under-sampled k-space data.

Main Results:

• The proposed algorithm successfully reconstructed MRI images from highly under-sampled data.
• A one-to-one correspondence between CT and MRI modalities was established.
• The algorithm demonstrated significantly superior performance compared to DL using MRI data alone, even with varying under-sampling factors and noise levels.

Conclusions:

• The novel joint CT-MRI reconstruction algorithm effectively enhances MRI image quality from under-sampled data.
• The EDGE method provides a good initial MRI estimation, improving subsequent DL reconstruction.
• This approach offers a promising direction for improving medical image reconstruction efficiency and accuracy.