Preview

PDF, English Download (13MB) | Terms of use

Abstract

Chemical exchange saturation transfer (CEST) magnetic resonance imaging (MRI) is an emerging, non-invasive molecular imaging technique. Particularly, the CEST-based amide proton transfer-weighted (APTw) contrast represents a valuable imaging biomarker for clinical assessment of brain cancers. However, the underlying molecular origin of the APTw contrast in vivo is still under debate. This thesis aims to expand the current knowledge about the APTw contrast mechanisms in the human brain at B0 = 3 T by developing a physically interpretable model (PIM) based on previously unexplored CEST signal features extracted by various machine learning methods. This PIM enabled the successful translation of interpretable CEST signals into their respective APTw contrast contributions not only as a black-box model but by explicitly exploiting physically relevant information. This novel approach allowed not merely (i) the identification of the isolated amide and exchange-relayed nuclear Overhauser effect (rNOE) contrasts as the dominating influences on the APTw contrast, in coherence with literature. But, more interestingly, (ii) this PIM also revealed significant dependencies of amide and rNOE contributions on changes in the controllable B1 and tissue-specific T1. Ultimately, the PIM allowed the identification of amide- and rNOE-driven sensitivity regimes of the APTw contrast, enabling an enhanced biophysical understanding of the CEST phenomenon in vivo, thus potentially improving the clinical assessment of brain cancers.