Exploring the Origins of the APTw Contrast in the Human Brain Using a Physically Interpretable Machine Learning Model

Motivation: Amide proton transfer-weighted (APTw) imaging is a promising biomarker for glioma assessment, however its origin remains debatable. Goal(s): This study seeks to improve the clinical assessment and interpretability of APTw imaging by investigating the mechanisms driving the contrast in the brain at 3T. Approach: The APTw contrast contributions were explored using a physically interpretable model. This model was created using data from 125 glioma patients, enabling the calculation of APTw contrast values based on more isolated CEST contrasts. Results: The analysis shows distinct sensitivity regimes for amide and rNOE contributions, expanding the current understanding of the APTw contrasts molecular origin in clinical imaging. Impact: This work introduces a physically interpretable model that decodes the APTw contrast mechanisms in the human brain, analyzing influences of amide and rNOE contrast. The model provides new insights into factors that impact APTw imaging, potentially enhancing the clinical interpretation.