A comparison of fitting algorithms for diffusion-weighted MRI data analysis using an intravoxel incoherent motion...

Area of Science:

• Medical Imaging
• Biophysics
• Computational Biology

Background:

• Diffusion-weighted magnetic resonance imaging (DW-MRI) is crucial for characterizing tissue microstructure.
• The intravoxel incoherent motion (IVIM) model provides insights into tissue perfusion and diffusion.
• Accurate fitting of the IVIM model is essential for reliable clinical interpretation.

Purpose of the Study:

• To develop and validate a modified VARiable PROjection (VARPRO) algorithm for IVIM model fitting.
• To specifically adapt the algorithm for diffusion-weighted MRI data in locally advanced rectal cancer (LARC).
• To compare the performance of the modified VARPRO algorithm against established and recent methods.

Main Methods:

• Implementation of a modified VARPRO algorithm for IVIM analysis.
• Comparison with non-linear least squares (NLLS) using Levenberg-Marquardt (LM) algorithm.
• Evaluation against two 'segmented' analysis algorithms.
• Performance assessment using the goodness-of-fit measure on DW-MRI data from 40 LARC patients.

Main Results:

• The modified VARPRO algorithm demonstrated superior performance compared to the LM algorithm in 64% of cases.
• The modified VARPRO algorithm outperformed 'segmented' analysis methods in 100% of cases.
• The proposed algorithm achieved a better fit for the IVIM model.

Conclusions:

• The modified VARPRO algorithm offers improved accuracy for fitting the IVIM model to LARC DW-MRI data.
• This enhanced fitting capability can potentially improve diagnostic and prognostic assessments in LARC.
• The algorithm provides a more robust alternative to existing methods for IVIM analysis in this clinical context.