Myelin water imaging data analysis in less than one minute

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Artificial Intelligence

Background:

• Myelin water imaging (MWI) is crucial for assessing white matter integrity.
• Accurate myelin water fraction (MWF) calculation is essential for MWI applications.
• Current MWI analysis methods can be time-consuming, limiting clinical feasibility.

Purpose of the Study:

• To develop a super-fast and easy-to-implement data analysis method for MWI using a deep learning neural network (NN).
• To accurately calculate the myelin water fraction (MWF) using the proposed NN algorithm.
• To significantly reduce the time required for MWF calculation to address clinical workflow limitations.

Main Methods:

• A neural network (NN) was constructed and trained on MWI data acquired using a 32-echo 3D gradient and spin echo (GRASE) sequence.
• Ground truth MWF values were generated using regularized non-negative least squares (NNLS) with stimulated echo corrections.
• The NN was trained and validated on data from multiple healthy and multiple sclerosis (MS) brains, as well as spinal cord data.

Main Results:

• The trained NN achieved whole-brain MWF map production in approximately 33 seconds without graphics card acceleration.
• No visual differences or regional biases were observed between the NN-derived and NNLS-derived MWF maps.
• Quantitative analysis showed excellent agreement between NN and NNLS methods, with R² > 0.98 and mean absolute error < 0.01.

Conclusions:

• The proposed NN algorithm dramatically reduces MWF calculation time to under one minute.
• This rapid analysis addresses a significant barrier to the clinical implementation of MWI.
• The NN provides an accurate and efficient method for MWF quantification, enhancing MWI's clinical potential.