Active Sampling for Accelerated MRI with Low-Rank Tensors.

Area of Science:

• Medical Imaging
• Biophysics
• Computational Imaging

Background:

• Magnetic Resonance Imaging (MRI) is crucial in medicine and biology but faces speed limitations in high-dimensional applications.
• Low-rank tensor models offer a solution for accelerated MRI through sparse sampling.
• Existing methods rely on predefined sampling, lacking adaptability.

Purpose of the Study:

• To introduce an active low-rank tensor model for accelerating Magnetic Resonance Imaging (MRI).
• To develop an active sampling strategy that leverages the benefits of low-rank tensor structures for efficient data acquisition.
• To address the speed constraints in high-dimensional MRI for broader practical utility.

Main Methods:

• Development of an active low-rank tensor model tailored for MRI.
• Implementation of an active sampling strategy utilizing a Query-by-Committee model.
• Validation using numerical experiments on a 3-D MRI dataset with Cartesian sampling.

Main Results:

• The proposed active low-rank tensor model effectively accelerates MRI acquisition.
• The active sampling method demonstrates superior performance compared to traditional approaches.
• Numerical experiments confirm the method's efficacy in enhancing imaging speed.

Conclusions:

• Active low-rank tensor modeling is a promising approach for fast MRI.
• The developed active sampling strategy significantly improves imaging efficiency.
• This method holds potential for advancing high-dimensional MRI applications.