Dictionary-Free MRI PERK: Parameter Estimation via Regression with Kernels.

Area of Science:

• Medical Imaging
• Computational Physics
• Machine Learning

Background:

• Quantitative magnetic resonance imaging (QMRI) parameter estimation is crucial for medical diagnostics.
• Current dictionary-based and iterative methods are computationally intensive, limiting their clinical application.
• There is a need for faster, accurate QMRI parameter estimation techniques.

Purpose of the Study:

• To introduce a novel, fast, and general method for dictionary-free QMRI parameter estimation.
• To demonstrate the efficiency and accuracy of the proposed method compared to existing techniques.
• To accelerate the process of quantitative parameter mapping in MRI.

Main Methods:

• Developed a method named Parameter Estimation via Regression with Kernels (PERK).
• PERK simulates parameter-measurement pairs using prior distributions and the nonlinear MR signal model.
• Employs kernel functions and convex optimization for nonlinear regression, followed by linear minimum mean-squared error regression.

Main Results:

• PERK achieves comparable T1 and T2 estimation accuracy to dictionary-based and iterative methods in white and gray matter.
• PERK demonstrates a speed improvement of at least 140x for T1 and T2 estimation.
• Potential for even greater acceleration (orders of magnitude) in complex, full-volume QMRI problems.

Conclusions:

• PERK offers a computationally efficient alternative for QMRI parameter estimation.
• The method's speed and accuracy make it suitable for accelerating clinical MRI workflows.
• PERK's generalizability suggests broad applicability in various QMRI applications.