QUANTIFYING WHITE MATTER HYPERINTENSITY AND BRAIN VOLUMES IN HETEROGENEOUS CLINICAL AND LOW-FIELD PORTABLE MRI

Area of Science:

• Neuroimaging
• Radiology
• Medical Image Analysis

Background:

• Brain atrophy and white matter hyperintensity (WMH) are key indicators of brain injury in cerebrovascular disease and multiple sclerosis.
• Current automated segmentation methods require high-resolution MRI with good signal-to-noise ratio (SNR), limiting their use in clinical and portable MRI (pMRI) settings.
• This limitation hinders large-scale tracking of neurodegenerative disease progression, particularly in underserved regions where pMRI is prevalent.

Purpose of the Study:

• To develop a robust method for segmenting WMH and 36 brain regions applicable to MRI scans of any resolution and contrast.
• To enable accurate quantification of brain atrophy and WMH progression using low-field pMRI data.
• To provide a universally applicable tool for neuroimaging analysis without the need for retraining.

Main Methods:

• A novel segmentation method was developed to process MRI scans regardless of resolution or contrast.
• The method was validated on eight public datasets and a private dataset with paired 3T and 64mT scans.
• The approach integrates seamlessly with existing neuroimaging software, such as FreeSurfer.

Main Results:

• The method successfully segmented white matter hyperintensity and 36 distinct brain regions across various MRI qualities.
• Strong correlations were observed between WMH (ρ=.85) and hippocampal volumes (ρ=.89) estimated from both high-field (3T) and low-field (64mT) scans.
• The technique demonstrated high accuracy and reliability across diverse datasets, including challenging pMRI data.

Conclusions:

• The developed method overcomes the limitations of existing techniques by enabling accurate segmentation of brain atrophy and WMH from low-field pMRI.
• This innovation facilitates large-scale tracking of neurodegenerative disease progression, especially in resource-limited settings.
• The publicly available tool enhances accessibility and applicability of advanced neuroimaging analysis in clinical practice and research.