Improved liver R2* mapping by pixel-wise curve fitting with adaptive neighborhood regularization

Area of Science:

• Medical Imaging
• Biophysics
• Computational Biology

Background:

• Liver R2* mapping using magnetic resonance imaging (MRI) is crucial for assessing hepatic iron overload.
• Standard R2* mapping techniques struggle with low signal-to-noise ratio (SNR) in serial MRI images, leading to inaccuracies.
• Improving the precision and accuracy of liver R2* quantification is essential for reliable clinical diagnosis.

Purpose of the Study:

• To develop and validate an advanced algorithm for liver R2* mapping that overcomes limitations of existing methods.
• To enhance the accuracy and precision of R2* measurements in the liver, particularly under low SNR conditions.
• To introduce adaptive neighborhood regularization into pixel-wise curve fitting for improved R2* quantification.

Main Methods:

• Proposed a novel algorithm: pixel-wise curve fitting with adaptive neighborhood regularization (PCANR).
• PCANR adaptively utilizes neighboring pixel information to regularize curve fitting, optimizing for interpixel signal similarity.
• Compared PCANR against conventional nonlinear least squares (NLS) and nonlocal means filter-based NLS algorithms using simulated, phantom, and in vivo data.

Main Results:

• PCANR generated R2* maps with substantially reduced noise and preserved fine anatomical structures compared to NLS methods.
• Quantitatively, PCANR demonstrated lower root mean square errors across various R2* values and SNR levels.
• PCANR exhibited superior accuracy and precision for high R2* values under low SNR conditions, outperforming NLS and nonlocal means filter-based NLS.

Conclusions:

• The PCANR algorithm effectively mitigates noise interference in liver R2* mapping.
• Enhanced measurement precision from PCANR is expected to improve the clinical assessment of hepatic iron.
• This method offers a promising advancement for quantitative MRI in clinical practice.