Constraints in estimating the proton density fat fraction

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Biomedical Engineering
• Medical Physics

Background:

• Proton density fat fraction (PDFF) is a key MRI biomarker for assessing tissue composition.
• Accurate PDFF quantification is crucial for diagnosing and monitoring various conditions.
• Current unconstrained estimation methods can be noisy and biased.

Purpose of the Study:

• To evaluate the impact of four physically motivated constraints on PDFF estimation.
• To assess the effectiveness of these constraints in reducing bias and standard deviation.
• To improve the accuracy and reliability of PDFF quantification using MRI.

Main Methods:

• Development of least squares approaches for PDFF quantification.
• Implementation of constraints: smooth fieldmap, smooth initial phase, nonnegative proton density, and moderate R₂^∗ values.
• Evaluation using numerical simulations and in vivo MRI data at 0.35 T.

Main Results:

• Unconstrained least squares estimation resulted in noisy and biased PDFF values.
• The evaluated constraints significantly reduced both bias and standard deviation.
• Constrained estimation demonstrated improved accuracy and precision compared to unconstrained methods.

Conclusions:

• Physically motivated constraints are effective in improving PDFF quantification accuracy.
• Constrained least squares approaches offer a more robust method for PDFF estimation in MRI.
• This study provides a foundation for more reliable fat fraction measurements in clinical applications.