A Modular Hybrid Framework for CEST MRI Development and Clinical Translation Using pTx-Pulseq Preparation With Open- and Closed-Source Readouts and ONNX Mapping.
Authors
Affiliations (8)
Affiliations (8)
- Institute of Neuroradiology, Uniklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany.
- Siemens Healthcare GmbH, Erlangen, Germany.
- Pattern Recognition Lab, Friedrich-Alexander-University Erlangen-Nürnberg, Erlangen, Germany.
- School of Biomedical Engineering, Tel Aviv University, Tel Aviv, Israel.
- Sagol School of Neuroscience, Tel Aviv University, Tel Aviv, Israel.
- Institute of Radiology, Uniklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Erlangen, Germany.
- Medical Physics in Radiology, German Cancer Research Center (DKFZ), Heidelberg, Germany.
- Department Artificial Intelligence in Biomedical Engineering (AIBE), Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany.
Abstract
Chemical exchange saturation transfer (CEST) MRI offers unique metabolic insights for imaging of cancer, stroke, or neurodegeneration. However, the rapid expansion of CEST preparation schemes and evaluation techniques creates a bottleneck; novel methods often remain complex prototypes requiring offline processing, preventing clinical adoption. To bridge this gap, we propose a hybrid sequence and reconstruction framework. This approach combines a flexible, pTx-enabled preparation block using the open-source Pulseq standard with a closed-source snapshot readout, and an interchangeable ONNX-based neural network for online post-processing. We validated this framework at 7 T using three distinct applications: deepWASABI for B0 and B1 prediction, a protein CEST approach for B1-corrected CEST mapping, and semi-solid MT MRF for quantifying magnetization transfer rates and semi-solid proton volume fractions. By utilizing closed-source readouts, we leverage advanced proprietary image reconstructions essential for high-resolution whole-brain coverage. Simultaneously, the integrated ONNX network processes contrast-weighted images directly at the scanner to generate parameter maps online. This hybrid architecture eliminates complex compilation steps, allowing deployment by simply copying Pulseq and ONNX files. By enabling push-button execution and immediate clinical data storage, this platform significantly accelerates the development, evaluation, and dissemination of advanced CEST MRI methods and forms a blueprint for scaling any magnetization-prepared MRI approach at any system.