Diagnostic value of deep learning–based renal virtual ASL sequences in CKD

Motivation: Arterial spin labeling (ASL), an MRI technique reflecting perfusion without contrast agents, can diagnose chronic kidney disease (CKD) but increases scan imaging cost and scan time. Goal(s): To explore the feasibility of generating virtual ASL from conventional MRI sequences. Approach: A model was developed using the pix2pix algorithm to generate virtual renal ASL sequences from T1 and T2-weighted imaging and diffusion-weighted imaging, and to measure renal blood flow values. Intraclass correlation coefficients, paired t tests, and receiver operating characteristic curves were used for data analysis. Results: Virtual ASL sequences could be generated through deep learning using conventional MRI sequences and could diagnose CKD. Impact: This virtual ASL technique enables non-invasive renal perfusion assessment based on conventional MRI sequences, potentially expanding access to perfusion imaging in CKD diagnosis and monitoring. Future studies can explore its application in treatment response prediction.