The AI-based model for dynamic MRI protocoling demonstrated strong diagnostic performance across multiple brain pathologies. In the initial stage, infarct detection reached 90.1% sensitivity and 92.8% specificity, while tumor and hemorrhage detection achieved sensitivities of 62.5% and 76.6%, respectively, with specificityies above 90%. A second-stage refinement increased tumor and hemorrhage sensitivities y to 75.5% and 84.4%, respectively, with only a modest reduction in specificity, demonstrating the model’s adaptability and robustness.
This AI-driven approach enables real-time adjustment of MRI scan sequences, improving workflow efficiency, standardization, and diagnostic confidence. Future integration with patient data and smart alerting systems could further advance personalized, intelligent MRI scanning for faster and more reliable care.