Advancing Z-Spectrum Analysis Protons (ZAP) Using Machine Learning with Random Forest Regression.

Motivation: To overcome long scan times in MRI's Z-Spectrum Analysis Protons (ZAP), impairing its clinical utility and patient comfort. Goal(s): This study aimed to refine ZAP, targeting a reduction in scan duration while maintaining high accuracy in proton exchange measurements. Approach: We applied Random Forest Regression to identify key offset frequencies, focusing on the most informative data and potentially reducing the scanning time. Results: Our approach successfully reduced scan times without compromising accuracy. The RFR model’s predictions aligned closely with traditional ZAP methods, indicating that fewer offset frequencies are needed for reliable data interpretation. Impact: This study may allow for targeted anatomical and disease-specific imaging with reduced scan times, potentially improving diagnostic accuracy and patient experience, and facilitating quicker, more focused clinical decisions.