Validation of non-invasive pressure drop quantification in an in-vitro stenosis phantom and patient cohort using DL-enhanced 4D Flow MRI

Motivation: Stenosis in major intracranial arteries is a key contributor to ischemic stroke, making precise, non-invasive trans-stenotic pressure assessment vital for determining stenosis severity. Goal(s): To validate a super-resolution-enhanced 4D Flow MRI technique for accurately estimating pressure drops across intracranial stenoses in both experimental and clinical settings. Approach: We applied a deep learning-based super-resolution tool, 4DFlowNet, to enhance 4D Flow MRI spatial resolution. Experimental validation in an in-vitro model and retrospective clinical cohort was performed using vWERP analysis. Results: Super-resolution MRI showed enhanced accuracy in pressure drop estimates, aligning closely with reference data and demonstrating improved flow visualization, particularly in severe stenoses. Impact: Deep learning-enhanced super-resolution 4D Flow MRI, coupled with physics-informed pressure estimation, enables precise, non-invasive quantification of pressure drops across intracranial stenoses, advancing clinical assessment of cerebrovascular disease and offering new insights into the hemodynamic impacts of arterial stenosis.