Info
Background:
Kinetic modeling enables estimation of perfluorooctyl bromide (PFOB) nanodroplet (ND) volume fraction (vB) from hyperpolarized (HP) 129Xe HyperCEST MR measurements, based on foundational contributions from Schröder, Bachert, Kunth, and others1-4. However, quantitative application of this framework requires prior knowledge of exchange and relaxation parameters, and the extent to which these parameters generalize across independently prepared ND formulations remains unknown.
Hypothesis/Objective:
In this work, we evaluate if a single calibrated parameter set can be applied across different ND batches and quantify the impact of global parameter transfer on vB estimation accuracy.
Methods:
ND surface-area-to-volume (Snp/Vnp) ratios were calculated for three different PFOB ND batches (0.0289, 0.0432, and 0.0349nm-1 for ND Batch 1, 2, and 3, respectively). In the kinetic model framework, Snp/Vnp directly influences the effective exchange rate between dissolved and ND-bound 129Xe and was therefore incorporated as a fixed geometric parameter during model calibration. HyperCEST spectroscopy was conducted using 200μL samples of PFOB NDs suspended in PBS. Measurements were acquired across seven ND concentrations corresponding to 2.735, 1.094, 0.547, 0.219, 0.055, 0.027, and 0.016mM PFOB.
HyperCEST preparation modules were incorporated into a simple pulse-acquire spectroscopy sequence. Saturation of HP 129Xe in PFOB was achieved through the use of sixteen 20ms three-lobe sinc pulses (310Hz bandwidth, 100μT peak B1) separated by 2ms spoiler gradients5. Three paired CEST OFF and CEST ON measurements were acquired sequentially for each sample. Saturation pulses were applied either off-resonance at 278ppm relative to dissolved 129Xe in water (179ppm) or on-resonance at 100ppm corresponding to 129Xe in PFOB NDs. Signal excitation and readout were performed at 179ppm with a minimum repetition time of 705ms. The three OFF/ON pairs were analyzed individually rather than averaged. Global exchange and relaxation parameters were determined independently for each ND batch using the previously developed kinetic modeling framework.
Results:
Using each batch-specific global parameter set, dataset-specific fits were subsequently performed to estimate vB for NDs and labeled-cell samples (Fig. 1). Batch-matched calibrations produced excellent agreement between fitted and true vB values for ND Batch 1 (R2 = 0.998), ND Batch 2 (R2 = 0.986), and ND Batch 3 labeled-cell datasets (R2 = 0.999). In contrast, application of mismatched parameter sets introduced systematic bias, indicating sensitivity to batch-specific parameter values. When datasets from ND Batch 1 and 2 were pooled and fit using a single combined calibration, fitted vB values remained strongly correlated with true vB (R² = 0.963). Although the combined calibration produced slightly reduced agreement relative to batch-specific fits, quantitative performance remained robust across both ND preparations, with the primary effect being an increased coefficient of variation in the fitted vB estimates.
Conclusion:
These findings demonstrate that accurate quantitative recovery of PFOB ND volume fraction is achievable, but estimation accuracy depends on agreement between the calibration dataset and ND formulation being analyzed. Batch-specific calibrations produced the most accurate vB estimates, whereas transfer of kinetic parameters between independently prepared ND formulations introduced systematic bias, suggesting that a single universal parameter set may not adequately describe HyperCEST behavior across all PFOB ND preparations.
Differences between batches are likely driven by variations in ND size distributions and corresponding surface-area-to-volume ratios, which directly influence exchange kinetics through the exchange term, kX = P*(Snp/Vnp)*vB. Because HyperCEST contrast arises from the combined effects of exchange, relaxation, and RF saturation, changes in ND geometry can alter the apparent kinetic behavior recovered during calibration.
Despite these formulation-dependent differences, pooled fitting across multiple ND batches maintained strong quantitative performance, suggesting that generalized calibrations may be feasible when applied to similar ND preparations. Together, these results establish practical limits on kinetic parameter transferability and highlight calibration strategy as an important consideration for quantitative HyperCEST MRI.
Image/Figure:
Click to view full size
Click to view full size
Image/Figure Caption:
Fig. 1. HyperCEST-derived vB estimation using batch-specific and combined global kinetic parameter calibrations. (a–b) Batch-matched calibrations produced close agreement between fitted and true vB, while mismatched parameter sets introduced systematic bias. (c) A combined global calibration preserved strong overall agreement across pooled datasets, although the fitted vB estimates exhibited increased coefficient of variation relative to batch-specific calibrations. (d) Batch 3 labeled-cell experiments also demonstrated excellent agreement between fitted and true vB. Dashed lines indicate the identity line (y = x).
Author
Collin J. Harlan
The University of Texas MD Anderson Cancer Center