Stochastic variational inference improves quantification of multiple timepoint arterial spin labelling perfusion MRI.

Area of Science:

• Neuroimaging
• Medical Physics
• Biomedical Engineering

Background:

• Arterial spin labelling (ASL) MRI enables non-invasive measurement of cerebral blood flow (CBF) and arterial transit time (ATT).
• ATT is an emerging biomarker for cerebrovascular health, but its accurate quantification is challenging due to low signal-to-noise ratio and complex tracer kinetics in ASL.
• Existing ASL quantification methods face limitations in accuracy and robustness.

Purpose of the Study:

• To introduce and validate a novel quantification method, SSVB, for multiple-timepoint ASL MRI.
• To address the limitations of current ASL data quantification techniques.
• To improve the accuracy and robustness of CBF and ATT measurements.

Main Methods:

• Development of the SSVB quantification method for ASL data.
• Validation using extensive simulations to assess accuracy and noise robustness.
• Application to high spatial resolution 3T ASL acquisition data from a cross-sectional cohort.

Main Results:

• Simulations demonstrated SSVB's superior accuracy and noise robustness compared to existing methods, especially for ATT estimation.
• SSVB produced less noisy parameter maps on high-resolution 3T ASL data.
• The method revealed greater variation in ATT within a cross-sectional cohort.

Conclusions:

• SSVB offers a more accurate and robust approach for quantifying ASL MRI data.
• This improved quantification enhances the reliability of CBF and ATT measurements.
• SSVB has the potential to advance the use of ASL MRI in assessing cerebrovascular health.