L-TGVN: Leveraging Longitudinal Priors for Personalized Rapid MRI

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Image Reconstruction

Background:

• Magnetic Resonance Imaging (MRI) offers superior soft-tissue contrast without ionizing radiation.
• Long MRI acquisition times lead to patient discomfort, increased costs, and limited scanner throughput.
• Accelerating MRI scans often requires incorporating prior knowledge to solve ill-posed reconstruction problems.

Purpose of the Study:

• To develop a method that leverages previous MRI scans to reconstruct current scans from undersampled data.
• To improve the efficiency and quality of MRI scans, particularly in longitudinal studies.
• To address challenges in using prior scans, such as temporal changes and misalignment.

Main Methods:

• Introduction of the Longitudinal Trust-Guided Variational Network (L-TGVN).
• L-TGVN uses prior scans as side information for reconstructing current scans from undersampled measurements.
• The network constrains the influence of prior scans to align with acquired measurements, without requiring explicit pre-registration and accommodating protocol drift.

Main Results:

• L-TGVN demonstrated consistent improvements in quantitative metrics compared to baseline methods.
• The method showed better preservation of fine structures at high acceleration rates.
• Evaluated against matched-capacity baselines, including prior-guided and non-longitudinal methods.

Conclusions:

• L-TGVN effectively reconstructs current MRI scans using longitudinal priors from previous scans.
• The method enhances image quality and structural preservation in accelerated MRI acquisition.
• L-TGVN offers a promising approach for efficient and high-quality longitudinal MRI studies.