Myocardial perfusion measurements with a deep learning-assisted cardiac arterial spin labeling (DeepCASL): towards validation by microsphere

Motivation: Cardiac arterial spin labeling (ASL) method is sensitive to noise (system and physiology), which may lead to inaccurate MBF measurement. Goal(s): A cardiac MRI arterial spin labeling method was developed with assistance of a deep learning networks (DeepCASL) to improve image quality and measurement accuracy. Approach: The performance of the DeepCASL method was evaluated in a canine model of coronary arterial disease by comparing and correlating with MBF determined by microsphere measurements. Results: The validation study revealed moderate to strong correlations in absolute myocardial blood flow values between MRI and microsphere reference methods. Impact: This new DeepCASL technique opens a door for clinical applications of noncontrast cardiac perfusion as a screen tool for reliable diagnosis of perfusion deficit in a variety of cardiomyopathy disorders.