A data-driven T₂ relaxation analysis approach for myelin water imaging: Spectrum analysis for multiple exponentials...

Area of Science:

• Biomedical Imaging and Spectroscopy
• Machine Learning in Medical Data Analysis

Background:

• Decomposition of multi-exponential decay data into T2 spectra is challenging for conventional methods like non-negative least squares (NNLS).
• Accurate T2 spectral analysis is crucial for quantitative imaging, particularly for myelin water fraction (MWF) estimation.

Purpose of the Study:

• To introduce a novel, data-driven analysis method, spectrum analysis for multiple exponentials via experimental condition oriented simulation (SAME-ECOS), for T2 spectral decomposition.
• To address the limitations of existing algorithms in resolving complex T2 decay signals and improve the accuracy of MWF quantification.

Main Methods:

• Developed SAME-ECOS, integrating resolution limit constraints with a machine learning neural network algorithm.
• Validated the SAME-ECOS workflow through theoretical derivation, computational simulations, and model training.
• Evaluated performance using simulated data and six in vivo human brain datasets, comparing against NNLS.

Main Results:

• SAME-ECOS demonstrated over 15% higher cosine similarity and more than 10% lower mean absolute error in MWF calculation compared to NNLS.
• The method exhibited superior robustness to noise in simulations and achieved distinct separation of myelin water peaks in in vivo data.
• SAME-ECOS processed whole-brain data approximately 30 times faster than NNLS, completing analysis in just 3 minutes.

Conclusions:

• SAME-ECOS provides more reliable T2 spectra and significantly reduces analysis time compared to NNLS.
• The enhanced speed and accuracy increase the feasibility of multi-component T2 decay analysis in clinical settings.
• This method offers a promising advancement for quantitative MRI applications.