Prospective Evaluation of an R2* Method for Assessing Liver Iron Concentration (LIC) Against FerriScan: Derivation of the Calibration Curve and Characterization of the Nature and Source of Uncertainty in the Relationship

Abstract

Background FerriScan is the method‐of‐choice for noninvasive liver iron concentration (LIC) quantification. However, it has a number of drawbacks including cost and expediency. Purpose/Hypothesis To characterize an R2*‐based MRI technique that may potentially be used as an alternative to FerriScan. This was accomplished through the derivation of a calibration curve that characterized the relationship between FerriScan‐derived LIC and R2*. The nature and source of uncertainty in this curve were investigated. It was hypothesized that the source of uncertainty is heterogeneity of LIC across the liver. Study Type Prospective. Subjects In all, 125 patients (69 women, 56 men) undergoing chelation treatment for iron overload prospectively underwent FerriScan and R2* MRI during the same exam. Field Strength/Sequence Pulse sequences included 2D multislice spin‐echo pulse for FerriScan, and a prototype 3D 6‐echo gradient echo acquisition for R2* mapping at 1.5T. Assessment A linear calibration curve was derived from the relationship between FerriScan‐derived LIC estimates and R2* through least‐squares fitting. Statistical Tests The nature of the uncertainty in the curve was characterized through tests of normality and homoscedasticity. The source of uncertainty was tested by comparing the magnitude of LIC variation over the FerriScan ROI to the observed uncertainty in the R2*‐derived LIC estimates. Results A linear relationship between logarithmically transformed FerriScan‐derived LIC and R2* (log{FerriScan‐derived LIC} = 1.029 log{R2*} – 3.822) was confirmed. Uncertainty was random, with a behaviour that was normal and homoscedastic. The source of uncertainty was confirmed as iron heterogeneity across the liver. The nontransformed calibration curve was: FerriScan‐derived LIC = 0.0266⋅R2*, with a constant coefficient‐of‐variation of 0.32. Data Conclusion FerriScan and R2* techniques were found to provide equivalent quantification of LIC in this study. Any difference in accuracy or precision was at a level lower than the uncertainty caused by variation in LIC over the liver. Level of Evidence: 1 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2019;49:1467–1474.

Fixing the MRI R2‐iron calibration in liver

“…Several large studies comparing liver iron concentration (LIC) by R2* and by Ferriscan ® R2 have identified substantial bias between R2* and R2 LIC estimates. [4][5][6] We postulated that the original Ferriscan R2 calibration overestimates LIC at high iron concentrations, exaggerating disagreements between the two techniques.…”

“…The calibration error is sufficient to completely explain the differences between LIC by R2* and Ferriscan ® R2 described in previously studies. [4][5][6] Importantly, any attempts to "calibrate" R2* or other MRI methods against Ferriscan ® need to account for this bias.…”

Histological Scores Validate the Accuracy of Hepatic Iron Load Measured by Signal Intensity Ratio and R2* Relaxometry MRI in Dialysis Patients

“…The second relates to the intrinsic limitations of MRI techniques for measurement of liver iron content. In addition to the importance of the skill of the radiologic team (both technician and radiologist) for the quality of the examinations, each technique has its own limitation: R2 relaxometry requires regular calibration of the apparatus by phantoms, is prone to respiratory motion artefacts and is influenced by steatosis; R2* is influenced both by iron and fat (explaining the development of multipeak spectral modelling methods) and its calibration is also controversial; and SIR requires a careful reproduction of Rennes University technical modalities and exhibits a loss of linearity at LIC values > 350 µmol/g (a situation rarely seen in haemodialysis-associated haemosiderosis) [15,19,[46][47][48]. Lastly, the variability of hepatic concentration in some liver diseases can also influence MRI measurement [42].…”

A novel gamma GLM approach to MRI relaxometry comparisons

“…Compared to traditional MRI imaging based on T 1 ‐ or T 2 ‐weighted signal intensity, MRI relaxometry is more robust for quantitative comparison as it is less sensitive to nuisance parameters such as coil positioning or scanner variabilities . Among other applications, relaxometry has been used to detect brain abnormalities and myocardial fibrosis or evaluate metabolic imaging, liver iron content (LIC), and contrast agent (CA) properties …”

“…This amounts to assuming that relaxometry data are lognormally distributed, rather than normally distributed. In fact, the log transform has been used occasionally for both linear mixed models and LIC calibration curves . Recently, the gamma distribution has also been used to account for the constant CV in neuroimaging relaxometry data .…”