4455
Clinical Pulsed CEST MRF Optimization using the Cramer-Rao Bound and Sequential Quadratic Programming1Department of Biomedical Engineering, Tel Aviv University, Tel Aviv, Israel, 2Institute of Neuroradiology, University Hospital Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Erlangen, Germany, 3Magnetic Resonance Center, Max-Planck-Institute for Biological Cybernetics, Tübingen, Germany, 4Department of Artificial Intelligence in Biomedical Engineering, Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Erlangen, Germany, 5Sagol School of Neuroscience, Tel Aviv University, Tel Aviv, Israel
Synopsis
Keywords: CEST / APT / NOE, CEST & MT, MRF
Motivation: The lack of a protocol optimization technique suitable for practical pulsed acquisition impedes the clinical translation of CEST MRF.
Goal(s): To develop a pulsed CEST MRF protocol optimization method, enabling improved parameter discrimination ability and accelerated acquisition.
Approach: The Cramer-Rao bound for variance assessment was employed on Bloch-McConnell-based simulated signals, followed by a numerical sequential quadratic programming optimization of MRF saturation pulse powers. Validation was performed using L-arginine phantoms for preclinical (7T) and clinical (3T) scanners.
Results: The proposed optimization approach resulted in significantly lower CEST-MRF reconstruction error (p<0.001) compared to baseline and a drastically short acquisition time (<20 s).
Impact: A pulsed CEST MRF optimization technique was developed, bridging the gap imposed by the lack of accurate analytical solutions and CEST optimization methods for practical clinical settings. The technique increases the translation potential of quantitative and rapid CEST imaging.
Introduction
Chemical exchange saturation transfer (CEST) MRI offers new opportunities for obtaining biological and clinical insights1. As routine clinical imaging is already composed of multiple and sequentially employed different protocols (reaching a total scan time of 30-60 min)2, CEST acquisition must be short, accurate, and optimized.Initially introduced for accelerated T1 and T2 mapping3-4, MR fingerprinting (MRF) has been successfully extended for CEST imaging5-6. CEST MRF enables the simultaneous quantification of multiple molecular properties, yet its discrimination ability is substantially affected by the acquisition protocol used6-7.
While several methods were recently proposed for the optimization of CEST MRF protocols8-9, they typically rely on the analytical solution of the Bloch-McConnell equations, which is not available or accurate for pulsed CEST acquisition (as commonly applied in clinical scanners).
The Cramer-Rao Bound (CRB) is a statistically derived concept for calculating estimation variance10, with demonstrated efficacy in optimizing quantitative T1/T2 protocols11; it was recently shown that the CRB reflects the parameter discrimination ability for a given CEST MRF protocol12. Here, we developed and validated a robust CRB-based CEST optimization technique, with applicability to any acquisition protocol.
Methods
As proof of concept, MRF imaging of L-arginine was optimized for preclinical continuous-wave (CW) and clinical pulsed acquisition, striving to improve the estimation accuracy of the proton exchange parameters and reduce the scan-time.Phantom preparation
Phantoms with L-arginine concentrations ranging from 25 to 200 mM at pH 4-6 were assembled and imaged at room temperature.
Baseline reference acquisition protocol
As the baseline CEST-MRF protocol, a previously reported acquisition schedule was used5,13, with a total saturation pulse length (Tsat) = 3 s (preclinical) / 2.5 s (clinical), recovery time (Trec) = 1 s, saturation pulse offset fixed at 3.0 ppm, and 30 random saturation pulse powers (B1). Preclinical imaging employed CW saturation with SE-EPI readout at 7T (Bruker, Germany), while clinical imaging utilized pulsed saturation (13 pulses, 100 ms pulses, 50% duty cycle) with a Snapshot CEST-EPI readout module14 at 3T (Prisma, Siemens).
Parameter optimization
Seeking an 86% acceleration in the acquisition, we redesigned the protocol to capture four parameter-encoding images instead of 30. All acquisition parameters were held constant except for the B1 values, which were randomly initialized. The optimization was performed using Scipy’s Sequential Quadratic Programming method15, employing CRB to minimize the cost function, defined as the sum of normalized CRB11 values for each estimated parameter (L-arginine concentration and proton exchange rate). The optimization process was repeated four times, generating an optimized protocol from a unique, randomly generated B1 pattern.
CEST-MRF dictionary generation and matching
Optimization and CRB calculation leveraged a fixed MRF dictionary, comprised of 69,299 entries, spanning the following parameter ranges: T1w=3.0s, T2w=[400-1500] ms, T1s=3.0s, T2s=40ms, fs=[10-120] mM, and ksw=[100-1400] Hz. A pulseq-CEST based16 C++ implemented simulator was used for numerically solving the Bloch–McConnell equations. The simulator was expanded with a Python interface for parallel execution capabilities, ensuring a practical run-time for the optimization process. Dot-product matching was used for reconstructing the quantitative parameter maps, and QUESP-based previously reported proton exchange rates were used for comparison5.
Results
For preclinical imaging, the optimization substantially decreased the mean absolute error (MAE) for L-arginine concentration from 85% to 19% (p<0.001, Fig. 1). While a decrease in proton exchange rate MAE was observed (from 21% to 16%), it was not statistically significant (p=0.17). Fig. 2 provides a representative example and comparison between the pre- and post-optimization quantitative CEST parameter maps.In the clinical scanner pulsed imaging case, after manual correction of substantial B0 and B1 inhomogeneities, the MAE for L-arginine concentration quantification significantly dropped from 42% to 19% (p<0.001, Fig. 3). Although quantification of the exchange rate with merely four raw MRF images proved challenging, Fig. 4 and Fig. 5 show the resulting parameter maps and their corresponding ground truth values.
Discussion
While both optimization experiments have significantly improved the parameter quantification accuracy, the improvement was more effective for preclinical imaging, potentially due to the improved SNR and B0/B1 homogeneity. Notably, the suggested approach resulted in a very short acquisition protocol, comprising merely four raw MRF images and reducing the scan time by 86% into less than 20 s. As optimizing the saturation pulse powers alone led to a drastic decrease in the MAE, future optimization of additional acquisition parameters is expected to provide even greater accuracy.Conclusion
A CEST MRF protocol optimization framework was developed, facilitating clinically relevant, reliable, and accurate quantitative imaging. Future studies could further leverage AI for faster optimization17 and a simultaneously improved reconstruction18,19.Acknowledgements
This project received funding from the European Research Council under the Horizon Europe program (grant agreement no. 101115639), the Ministry of Innovation, Science and Technology, Israel, and a grant from the Tel Aviv University Center for AI and Data Science (TAD).
References
1. Liu, G., Song, X., et al. “Nuts and bolts of chemical exchange saturation transfer MRI.” NMR in Biomedicine. 2013;26(7):810-828. doi: 10.1002/nbm.2899.
2. Vladimirov, N., Perlman, O. Molecular MRI-Based Monitoring of Cancer Immunotherapy Treatment Response. Int. J. Mol. Sci. 2023;24:3151. https://doi.org/10.3390/ijms24043151
3. Ma, D. et al. “Magnetic resonance fingerprinting.” Nature. 2013;495(7440): Art. no. 7440. doi: 10.1038/nature11971.
4. Poorman, M. E. et al. “Magnetic resonance fingerprinting Part 1: Potential uses, current challenges, and recommendations.” Journal of Magnetic Resonance Imaging. 2020;51(3):675-692. doi: 10.1002/jmri.26836.
5. Cohen, O., et al. “Rapid and quantitative chemical exchange saturation transfer (CEST) imaging with magnetic resonance fingerprinting (MRF).” Magn. Reson. Med. 2018;80:2449-2463.
6. Perlman, O., Farrar, C. T., et al. "MR fingerprinting for semisolid magnetization transfer and chemical exchange saturation transfer quantification." NMR in Biomedicine. 2023;36(6):e4710.
7. Perlman, O., et al. “CEST MR‐fingerprinting: practical considerations and insights for acquisition schedule design and improved reconstruction.” Magnetic Resonance in Medicine. 2020;83(2):462-478.
8. Perlman, O., et al. “An end‐to‐end AI‐based framework for automated discovery of rapid CEST/MT MRI acquisition protocols and molecular parameter quantification (AutoCEST).” Magnetic Resonance in Medicine. 2022;87(6):2792-2810.
9. Kang, B., et al. “Learning‐based optimization of acquisition schedule for magnetization transfer contrast MR fingerprinting.” NMR in Biomedicine. 2022;35(5):e4662.
10. Kay, S. M. “Fundamentals of Statistical Signal Processing, Volume I: Estimation Theory." 1st ed. Upper Saddle River, NJ: Pearson; 1993.
11. Zhao, B. et al. “Optimal experiment design for magnetic resonance fingerprinting: Cramer-Rao bound meets spin dynamics.” IEEE Trans. Med. Imaging. 2018;38:844-861.
12. Liu, J. et al. “Encoding capability prediction of acquisition schedules in CEST MR fingerprinting for pH quantification.” Magnetic Resonance in Medicine. 2022;87(4):2044-2052. doi: 10.1002/mrm.29074.
13. Weigand‐Whittier, J. et al. “Accelerated and quantitative three‐dimensional molecular MRI using a generative adversarial network.” Magnetic Resonance in Medicine. 2023;89(5):1901-1914.
14. Mueller, S., et al. “Whole brain snapshot CEST at 3T using 3D‐EPI: Aiming for speed, volume, and homogeneity.” Magnetic Resonance in Medicine. 2020;84(5):2469-2483.
15. Virtanen, P. et al. “SciPy 1.0: Fundamental Algorithms for Scientific Computing in Python.” Nature Methods. 2020;17(3):261-272.
16. Herz, K., Mueller, S., Perlman, O., et al. “Pulseq-CEST: Towards multi-site multi-vendor compatibility and reproducibility of CEST experiments using an open-source sequence standard.” Magn Reson Med. doi: 10.1002/mrm.28825
17. Nagar, D., et al. “Dynamic and rapid deep synthesis of chemical exchange saturation transfer and semisolid magnetization transfer MRI signals.” Sci Rep. 2023;13:18291. https://doi.org/10.1038/s41598-023-45548-8
18. Cohen, O., et al. “CEST MR fingerprinting (CEST-MRF) for brain tumor quantification using EPI readout and deep learning reconstruction.” Magn Reson Med. 2023;89:233-249. doi:10.1002/mrm.29448
19. Perlman, O., et al. “Quantitative imaging of apoptosis following oncolytic virotherapy by magnetic resonance fingerprinting aided by deep learning.” Nature Biomedical Engineering. 2022;6(5):648-657.
Figures
