Introduction

Measurement of hepatic iron overload is an important clinical problem in patients with hemosiderosis from chronic transfusions, iron hyperabsorption syndromes, and liver disease¹. R2* or R2 relaxation values are calculated from multiple TE gradient echo or spin-echo sequences, but both fail at high LIC because signal is extinguished over most TEs¹. In this situation, however, there is often sufficient signal in the first echo to characterize the ratio between liver and muscle darkening. We postulated that we could formulate a generalizable framework for LIC estimation at any TE and field strength using basic principles of the signal decay equation.

Methods

All subjects provided informed consent under an IRB-approved protocol(CCI#14-00034). All subjects underwent assessment of liver R2* at 1.5T (Philips Achieva) and 3.0T (Philips Ingenia) scanners in close temporal proximity (5±11 days) using product multi-echo, gradient echo pulse sequences. Reference LIC from the 1.5T scan was calculated using previously published methods¹ and calibration². Signal intensity ratios (SIRs) were calculated from a region of interest(ROI) placed just inside the major curvature of the liver, and two ~4 cm² circular ROI’s place in the paraspinous muscles in the same imaging plane (Figure 1). The liver-muscle SIR can be written as shown in Figure 2, Eq. [1], where PD is the proton density, TR is the repetition time, TE is the echo time, T1 is the longitudinal relaxation rate, θ is the flip angle, R2* is the transverse relaxation rate and the subscripts L and M designate “liver” and “muscle” specific values, respectively. Taking the log, collecting terms, and substituting delta R2* = k(B₀/B_ref)LIC yields Equation [2], where the PD and T1 terms have been collected in a parameter ‘A’. Once A and k have been estimated, LIC_SIR estimates at any field strength can be calculated using Equation [3].

Results

Figure 3 demonstrates log(SIR) plotted against the product of TE, B₀/1.5T, and LIC and linear fits at 1.5T and 3.0T. Neither the slope or intercept were different across field strength, resulting in a pooled k and log(A) of -0.034 and 0.4209, respectively, r=-0.965, p<0.0001. Figure 4 demonstrates the LIC_SIR for both 1.5T and 3.0T values plotted against the reference, 1.5T LIC_R2* value. There is excellent agreement across the entire dynamic range of iron overload (r² = 0.94). Bland-Altman analysis is shown in Figure 5. LIC_SIR and LIC_R2* estimates were unbiased with means and standard deviations of 0.8 ± 3.2 mg/g (1.5T), and -0.5 ± 3.7 (3.0T) mg/g, respectively.

Discussion

LIC_R2* more accurate than LIC_SIR at low to moderate iron overload because it controls for signal variations caused by T1 decay, B1+ inhomogeneity, and nonuniform coil sensitivities. However, LIC_SIR estimates offer reasonable accuracy when signal loss precludes liver R2* measurement. In Figure 2, for example, LIC_SIR estimates at 3T were accurate up to 40 mg/g, more than double the typical R2*_LIC dynamic range achievable at 3T³. To provide context, roughly 6% of our routine clinical R2* assessments exceed the dynamic range at 1.5T and 20% fail at 3T. With more 3T magnets being installed, the proposed simple mechanism to overcome the dynamic range limitations could have important consequences.

The use of SIR for LIC estimation has a long history, but previous applications have placed stringent restrictions on imaging hardware (such avoiding phased-array coils) and scan parameters to improve LIC prediction for mild to moderate iron overload⁴. Most of the uncertainty in equation 2 (and Figures 4, 5), arise from the subject and scanner dependent factors in the ‘A’ term. This uncertainty is largely LIC independent (see Figure 5), becoming relatively less important at higher iron concentrations. That is, a standard-deviation of 3.2-3.7 mg/g is much more clinically acceptable at LICs greater than 20 mg/g, than below it. This allows “rescue” LIC calculation even with nonoptimized imaging parameters. In this study, default surface coil intensity compensation was used (CLEAR option) but there was no attempt to correct for the known, large, fluctuations in B1+ inhomogeneity⁵ from our single transmit systems. We did examine bootstrap corrections for T1 decay⁶, but inclusion did not improve LIC estimation (not shown).

LIC_SIR estimation could even be used to retrospectively estimate LIC in abdominal MRI studies that never acquired quantitative relaxometry. For example, most imaging protocols include in-phase/out-of-phase gradient echo imaging that could generate LIC estimates suitable for population-based clinical research.

Conclusion

In this paper, we extend a decades-old approach of LIC estimation to derive a field and parameter indifferent calibration that can extend the dynamic range of R2* (or R2) acquisitions, complementing multiecho relaxometry. We are currently performing cross vender and cross-sequence validation.

Acknowledgements

This work supported by the National Institute for Diabetes, Digestive and Kidney Diseases (1R01DK097115-01A1), the National Center for Research Resources (UL1 TR001855-02), the Saban Research Institute, Agios Pharmaceuticals, and Philips Healthcare (Research Support In Kind)