[Paper Review] CEST MR fingerprinting (CEST-MRF) for Brain Tumor Quantification Using EPI Readout and Deep Learning Reconstruction

Ouri Cohen, Victoria Yu+5 more|arXiv (Cornell University)|Aug 18, 2021

Advanced MRI Techniques and Applications39 references4 citations

TL;DR

This study presents a clinical CEST-MRI fingerprinting method using EPI readout and deep learning reconstruction (DRONE) for rapid, accurate quantification of brain tumors in under 2 minutes. The approach achieves high reproducibility (CCC ≥ 0.98) and detects significant CEST parameter differences in tumor regions versus contralateral tissue, demonstrating clinical potential for non-invasive tumor characterization.

ABSTRACT

$ extbf{Purpose}$: To develop a clinical CEST MR fingerprinting (CEST-MRF) method for brain tumor quantification using EPI acquisition and deep learning reconstruction. $ extbf{Methods}$: A CEST-MRF pulse sequence originally designed for animal imaging was modified to conform to hardware limits on clinical scanners while keeping scan time $\leq$ 2 minutes. Quantitative MRF reconstruction was performed using a deep reconstruction network (DRONE) to yield the water relaxation and chemical exchange parameters. The feasibility of the 6 parameter DRONE reconstruction was tested in simulations in a digital brain phantom. A healthy subject was scanned with the CEST-MRF sequence, conventional MRF and CEST sequences for comparison. Reproducibility was assessed via test-retest experiments and the concordance correlation coefficient (CCC) calculated for white matter (WM) and grey matter (GM). The clinical utility of CEST-MRF was demonstrated in 4 patients with brain metastases in comparison to standard clinical imaging sequences. Tumors were segmented into edema, solid core and necrotic core regions and the CEST-MRF values compared to the contra-lateral side. $ extbf{Results}$: The DRONE reconstruction of the digital phantom yielded a normalized RMS error of $\leq$ 7% for all parameters. The CEST-MRF parameters were in good agreement with those from conventional MRF and CEST sequences and previous studies. The mean CCC for all 6 parameters was 0.98$\pm$0.01 in WM and 0.98$\pm$0.02 in GM. The CEST-MRF values in nearly all tumor regions were significantly different (P=0.05) from each other and the contra-lateral side. $ extbf{Conclusion}$: Combination of EPI readout and deep learning reconstruction enabled fast, accurate and reproducible CEST-MRF in brain tumors.

Motivation & Objective

• To develop a clinically feasible CEST-MRI fingerprinting method compatible with standard MRI scanners.
• To enable rapid acquisition using EPI readout while maintaining diagnostic accuracy.
• To improve reconstruction accuracy and speed using a deep learning network (DRONE).
• To validate the method’s reproducibility and clinical utility in brain tumor patients.
• To compare CEST-MRF parameters with conventional MRF and CEST sequences for consistency.

Proposed method

• Modified a pre-existing animal-focused CEST-MRF pulse sequence to comply with clinical scanner hardware constraints.
• Employed echo-planar imaging (EPI) for fast k-space sampling to keep scan time ≤ 2 minutes.
• Applied the DRONE deep learning reconstruction network to reconstruct six quantitative parameters: T1, T2, CEST, and three exchange rates.
• Validated the reconstruction pipeline using a digital brain phantom with simulated tissue properties.
• Conducted test-retest experiments in a healthy subject to assess reproducibility using concordance correlation coefficient (CCC).
• Performed clinical scans in four patients with brain metastases, segmenting tumors into edema, solid core, and necrotic core for regional analysis.

Experimental results

Research questions

• RQ1 Can CEST-MRF be adapted for clinical use on standard MRI scanners with limited hardware capabilities?
• RQ2 Does EPI-based acquisition enable sub-2-minute scan times without compromising accuracy?
• RQ3 Can deep learning reconstruction (DRONE) achieve high-fidelity recovery of six quantitative parameters from undersampled data?
• RQ4 How reproducible are CEST-MRF parameters across repeated scans in healthy brain tissue?
• RQ5 Are CEST-MRF values significantly different in tumor regions compared to the contralateral normal-appearing tissue?

Key findings

• The DRONE reconstruction achieved a normalized RMS error ≤ 7% for all six parameters in the digital brain phantom, confirming high reconstruction accuracy.
• CEST-MRF parameters showed strong agreement with those from conventional MRF and CEST sequences, validating methodological consistency.
• Test-retest reproducibility was excellent, with a mean CCC of 0.98 ± 0.01 in white matter and 0.98 ± 0.02 in grey matter.
• In all four patients, CEST-MRF values in tumor regions (edema, solid core, necrotic core) were significantly different (p = 0.05) from the contralateral side.
• The method successfully differentiated tumor subregions, demonstrating sensitivity to microenvironmental changes in brain metastases.
• The combination of EPI readout and DRONE reconstruction enabled fast, accurate, and reproducible quantification suitable for clinical brain tumor assessment.