Artificial Intelligence to Eliminate the Need for Gadolinium-Based Contrast Agents in Brain Perfusion Weighted Imaging

Motivation: Acute ischemic stroke (AIS) requires rapid intervention, but current methods for quantifying cerebral hemodynamic parameters rely on contrast agents, which are time-consuming and carry patient risks. Non-invasive alternatives are needed. Goal(s): This study aims to use AI to quantify CBV, CBF, and MTT from non-contrast-enhanced imaging, providing a safer and faster alternative to contrast-based methods. Approach: Two AI models (one-stage and two-stage) were trained on multimodal imaging from 120 subjects and evaluated through multiple metrics. Results: The two-stage model outperformed the one-stage model, demonstrating higher consistency in parameter quantification for both whole-brain and lesion regions. Impact: AI-generated hemodynamic parameters from non-contrast imaging offer significant clinical benefits, including reduced costs, faster acquisition, and improved patient safety. This approach could streamline stroke diagnosis and improve healthcare accessibility.