Machine Learning Based Approach for Partial Volume Corrected Cerebral Blood Flow and Arterial Transit Time Mapping

Motivation: To reduce a scan time of multi-PLD PCASL imaging using a convolutional neural network (CNN) for robust estimation of partial volume (PV)-corrected ATT and CBF maps. Goal(s): To develop a CNN to predict PV-corrected ATT and CBF from fewer PLDs, ensuring minimal accuracy loss. Approach: Trained and validated a CNN on multi-PLD ASL data from 48 subjects, comparing its performance with a standard method. Results: The CNN achieved low mean average errors, suggesting reduced PLD count does not significantly affect ATT and CBF estimation accuracy. Impact: The study’s CNN reduces MRI scan times while accurately estimating brain hemodynamic parameters, such as cerebral blood flow and arterial transit time, enhancing patient comfort and diagnostic efficiency, potentially transforming cerebrovascular disease monitoring and advancing AI integration in medical imaging.