Arterial spin labeling cerebral blood flow quantification from quantitative transport mapping based on multiscale...

Area of Science:

• Neuroimaging
• Medical Physics
• Computational Fluid Dynamics

Background:

• Traditional kinetic modeling (TKM) for arterial spin labeling (ASL) relies on a global arterial input function (AIF) and uniform flow assumptions.
• These assumptions lead to systematic errors, overestimating arterial input and underestimating cerebral blood flow (CBF).
• Quantitative transport mapping (QTM) offers an alternative, overcoming AIF-related limitations.

Purpose of the Study:

• To introduce QTM deep neural network (QTMnet) for accurate CBF quantification from ASL data.
• To evaluate QTMnet's performance against established methods and ground truth in both simulated and in vivo scenarios.

Main Methods:

• QTMnet was trained using synthesized multi-delay ASL data derived from multiscale vascular fluid mechanics simulations.
• Validation involved comparing QTMnet's CBF quantification against in silico ground truth and phase-contrast (PC) MRI measured total brain flow in vivo.
• Performance was benchmarked against the traditional kinetic model (TKM).

Main Results:

• QTMnet significantly reduced CBF underestimation compared to TKM, by approximately 3-4 times.
• In silico validation showed TKM underestimation biases of -12.80/-9.73 mL/100g/min, reduced to -3.96 mL/100g/min by QTMnet.
• In vivo validation against PC flow demonstrated TKM underestimation biases of -16.0±3.1/-13.5±3.3 mL/100g/min, improved to -3.31±2.7 mL/100g/min by QTMnet.

Conclusions:

• QTMnet represents a promising advancement for CBF quantification using ASL.
• The fluid-mechanics-based QTMnet demonstrates superior accuracy compared to TKM in both simulated and in vivo ASL data.
• QTMnet offers a more precise method for assessing cerebral blood flow from ASL imaging.