DEEP FACTOR MODEL: A NOVEL APPROACH FOR MOTION COMPENSATED MULTI-DIMENSIONAL MRI.

Area of Science:

• Medical Imaging
• Magnetic Resonance Imaging
• Computational Imaging

Background:

• Quantitative parameter mapping in MRI, such as MR fingerprinting (MRF), captures magnetization evolution over time.
• Current methods often require extensive data acquisition, leading to long scan times, especially for high-resolution 3D applications.
• Subject motion during scans can degrade image quality and complicate analysis.

Purpose of the Study:

• To introduce the Deep Factor Model (DFM), a novel approach for efficient representation of multi-contrast MRI time series.
• To enable faster 3D high-resolution quantitative parameter mapping by allowing highly undersampled image acquisition.
• To enhance the robustness of MRI scans against subject motion through integrated motion estimation and compensation.

Main Methods:

• Developed a Deep Factor Model (DFM) for efficient representation of dynamic MRI data.
• Integrated motion estimation and compensation algorithms within the DFM framework.
• Applied the DFM to multi-contrast, 3D, high-resolution MRI data acquisition scenarios.

Main Results:

• The DFM provides a significantly more efficient representation of multi-contrast MR image time series compared to existing methods.
• The efficient representation facilitates highly undersampled data acquisition, leading to reduced overall scan times.
• The integrated motion compensation demonstrated robustness against subject motion, preserving image quality.

Conclusions:

• The Deep Factor Model (DFM) offers a powerful and efficient method for quantitative parameter mapping in MRI.
• DFM enables accelerated 3D high-resolution imaging by reducing undersampling artifacts and scan duration.
• The integrated motion correction makes DFM a robust technique for clinical and research applications where patient movement is a concern.