Improving Liver Iron Estimates with a Muscle-based Proton Density Estimate
Eamon K Doyle, MS1, Andrew Powell, MD2,3, and John C Wood, MD, PhD4,5

1Biomedical Engineering, University of Southern California, Sierra Madre, CA, United States, 2Cardiology, Boston Children's Hospital, Boston, MA, United States, 3Pediatrics, Harvard School of Medicine, Boston, MA, United States, 4Cardiology, Children's Hospital of Los Angeles, Los Angeles, CA, United States, 5Biomedical Engineering, University of Southern California, Los Angeles, CA, United States

Synopsis

CPMG-based R2 (1/T2) estimates are traditionally insensitive to tissue iron load. We show that the application of a T1-corrected, skeletal muscle-based proton density constraint increases the sensitivity of R2 for iron quantitation in phantoms and human subjects. This method leads to a fundamentally different R2-LIC (liver iron concentration) calibration curve than has previously been applied to CPMG fit data.

Introduction

MRI has proven to be a useful modality to noninvasively diagnose and quantify iron overload disorders.[1] R2 (1/T2) is commonly estimated by fitting a decaying exponential signal model to multiple single echo or MESE image series. Existing calibration curves show the relationship between R2 and liver iron concentration (LIC) to be more shallow than the R2* (1/T2*) vs LIC relationship[2], particularly for multiple-echo, spin-echo (MESE) acquisitions. In this work, we use Monte Carlo simulation to explore the decreased R2 sensitivity with iron load in MESE scans[3]. We also demonstrate that a muscle-based proton density estimate can increase the sensitivity of MESE LIC estimation in patient and simulation data.

Methods

Clinically indicated MRI iron assessments were performed on 37 iron-loaded human subjects on a GE 1.5T Signa Twinspeed magnet. Liver R2* was measured in a single mid-hepatic slice using a multiple-echo gradient echo sequence with echo times from 1.5 – 21.4 ms.; R2* was used as the reference standard for LIC estimation. Liver R2 was measured in a single 8mm, mid-hepatic slice using CPMG multiple-echo spin echo sequence with TE/TR = [6.5,13,19.5,26,32.5,39,45.5,52]/246ms, BW=4111Hz/px, NEX=1, and matrix=64x64. Regions of interest (ROI) in the liver and skeletal muscle were segmented by a research assistant with 5 years of experience segmenting abdominal MR images.

A CPMG spin echo pulse sequence was simulated using a previously developed and validated framework[3] including a tissue modeling engine[4], proton diffusion generator, and complete pulse sequence and Bloch simulator.[5] MRI signals for tissue iron loads were simulated over a range of liver iron loads from 1-50 mg/g FE/dry tissue. Relevant sequence parameters were matched to clinical spin echo assessments.

Fitting was performed using a monoexponential+constant signal model in MATLAB. Data was fit with and without a proton density constraint (PDE) to compare fits to LIC. Human subject PDEs bounds were estimated to be $$$\pm10%$$$ of T1-corrected mean signal intensity in the muscle ROI; simulation PDE boundaries were set to $$$\pm10%$$$ of an a priori known proton density. Three fit iterations were used to solve for LIC-dependent T1 values; muscle T1 was assumed to be 1008 ms[6].

Results

Figure 1 demonstrates R2 as a function of liver iron concentration in 27 patients (solid dots). R2 rises very slowly with iron concentration, unlike the calibration curve observed in single spin-echo experiments[2]. Simulation data (Figure 1, solid line) accurately predicts the shallow R2-LIC relationship when the fitting is limited to data at each of the echo times. When simulation data includes an estimate of proton density, the sensitivity of R2 to iron increases threefold (Figure 2, solid line). Inclusion of an image-derived PDE into the patient R2 estimates also increases sensitivity with iron (Figure 2, solid dots). There is greater scatter in the PDE-constrained patient data than observed in unconstrained fits but the iron sensitivity is also increased: all patient R2 estimates for LICs over 4 mg/g were greater for the PDE constrained fits than the unconstrained fits.

Discussion

Iron-mediated transverse relaxation demonstrates a theoretically non-exponential signal model[7] similar to a biexponential curve when measured with MESE sequences. At echo times used in most clinical imaging studies, the rapid decay component is not captured, leaving only the slow decay component to be acquired. This leads to characteristically flat calibration curves seen with MESE T2-LIC estimates. LIC proton density estimates have traditionally been considered nuisance parameters and are often ignored. However, we find that a constrained proton density estimate helps to restore the fast decay information that is normally lost and leads to a fundamentally different R2-LIC calibration with greater sensitivity.

The patient results demonstrated are part of a larger dataset currently undergoing analysis. Future work includes applying additional signal models such as nonexponential decay[7] to human and simulation data and completing evaluation of data from patients with up to 40 mg/g to characterize the effects of a PDE over the entire clinical range of iron loads. We are currently characterizing liver T1 values in iron overloaded subjects and will refine our iterative liver proton density estimator according to the results. We expect these refinements to improve the scatter and sensitivity seen in the PDE-constrained patient R2 fits, moving them closer to the simulated curve in Figure 2.

Conclusion

Inclusion of a proton density constraint estimated from skeletal muscle leads to increased sensitivity of R2 estimates with respect to liver iron load. The resulting R2-LIC calibration is fundamentally different due to the non-exponential nature of iron-mediated R2 relaxation in MESE sequences. This method shows promise to improve accuracy, sensitivity, and applicability of MESE in R2 estimation.

Acknowledgements

This work is supported by the National Institute of Health, National Institute of Diabetes and Digestive and Kidney Diseases, Grant R01-DK097115. Computation for the work described in this paper was supported by the University of Southern California’s Center for High-Performance Computing (hpc.usc.edu).

References

[1] J. C. Wood, C. Enriquez, N. Ghugre, J. M. Tyzka, S. Carson, M. D. Nelson, and T. D. Coates, “MRI R2 and R2* mapping accurately estimates hepatic iron concentration in transfusion-dependent thalassemia and sickle cell disease patients,” Blood, vol. 106, no. 4, pp. 1460–1465, Aug. 2005.

[2] T. G. S. Pierre, P. R. Clark, W. Chua-anusorn, A. J. Fleming, G. P. Jeffrey, J. K. Olynyk, P. Pootrakul, E. Robins, and R. Lindeman, “Noninvasive measurement and imaging of liver iron concentrations using proton magnetic resonance,” Blood, vol. 105, no. 2, pp. 855–861, Jan. 2005.

[3] N. R. Ghugre and J. C. Wood, “Relaxivity-iron calibration in hepatic iron overload: Probing underlying biophysical mechanisms using a Monte Carlo model,” Magn. Reson. Med. Off. J. Soc. Magn. Reson. Med. Soc. Magn. Reson. Med., vol. 65, no. 3, pp. 837–847, Mar. 2011.

[4] N. R. Ghugre, I. Gonzalez-Gomez, H. Shimada, T. D. Coates, and J. C. Wood, “Quantitative analysis and modelling of hepatic iron stores using stereology and spatial statistics,” J. Microsc., vol. 238, no. 3, pp. 265–274, Jun. 2010.

[5] E. K. Doyle and Wood, John C, “Effects of B1+ Inhomogeneity on Liver Iron Estimates in MRI,” presented at the BMES, Tampa, FL, USA, 2015.

[6] G. J. Stanisz, E. E. Odrobina, J. Pun, M. Escaravage, S. J. Graham, M. J. Bronskill, and R. M. Henkelman, “T1, T2 relaxation and magnetization transfer in tissue at 3T,” Magn. Reson. Med. Off. J. Soc. Magn. Reson. Med. Soc. Magn. Reson. Med., vol. 54, no. 3, pp. 507–512, Sep. 2005.

[7] J. H. Jensen and R. Chandra, “Theory of nonexponential NMR signal decay in liver with iron overload or superparamagnetic iron oxide particles,” Magn. Reson. Med., vol. 47, no. 6, pp. 1131–1138, Jun. 2002.

Figures

Figure 1 - ROI-averaged human R2 estimates are plotted against R2*-based LIC estimates. Simulation R2 estimates are plotted against known sample LIC. Unconstrained fits show a shallow relationship compared to iron load.

Figure 2 - ROI-averaged human R2 estimates are plotted against R2*-based LIC estimates. Simulation R2 estimates are plotted against known sample LIC. The sensitivity of simulation fits is increased when a proton density estimate is used. Scatter in patient R2 estimates is elevated compared to unconstrained fits.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
3859