Comparison of hepatic iron quantification by MRI between 1.5 and 3 T, using different methods and conversion formulas Yves Gandon, Thi Hien Trang Chau, Paul Borde, Elise Bannier, Thibault Lapotre, Mustapha Azahaf and Edouard Bardou-Jacquet Eur Radiol (2025). Published 27 September 2025. https://doi.org/10.1007/s00330-025-12027-9
The study compares different MRI methods for the assessment of liver iron concentration. The non-invasive assessment of liver iron and fat accumulation is of increasing importance as a rising prevalence of MASLD (metabolic associated steatotic liver disease) is seen. In MASLD not only fat, but also iron accumulates in the liver leading to severe liver damage over time. MRI is considered a powerful, accurate, and rapid tool for the simultaneous quantification of both liver fat and iron, having largely replaced invasive biopsy for routine clinical practice. In particular, R2* relaxometry using a multi-echo gradient recalled echo (MR GRE) sequence is recommended for simultaneous liver fat and iron quantification. R2*, unlike proton density fat fraction (PDFF) used for fat quantification, is proportional to the magnetic field strength. Furthermore, clinicians are more familiar with the liver iron concentration (LIC) value, thus R2* needs to be converted into LIC. Several conversion formulas have been proposed differing between field strength and between European and American studies. Currently, major MRI vendors offer commercial sequences with specialized algorithms producing whole-liver PDFF and R2* maps. Alternatively, non-specific chemical shift-encoded 2D or 3D ME GRE sequences, analyzed with independent software, can also be used for quantifying liver PDFF and R2*. The purpose of this study was to compare LIC results derived from 1.5T and 3T MRI scanners using vendor-specific and non-specific sequences and using “European” and “American” conversion formulas. In this prospective, two-center study 78 patients underwent two consecutive MRI exams, one on a 1.5T MR system and one on a 3T MR system for liver iron and fat quantification. On both systems a non-specific ME GRE sequence according to MRQuantif specifications and a vendor specific sequence was performed (IDEAL-IQ from GE Healthcare and LiverLab from Siemens Healthineers). Proton Density Fat Fraction (PDFF) and R2* values were obtained from vendor sequences via map reconstruction and from non-specific sequences using MRQuantif software. LIC was derived from R2* using European (Garbowski (1), d'Assignies (2)) or American (Hernando (3)) calibrations. Data was analysed using linear correlation regression and Bland-Altman plots to evaluate agreement and bias between LIC values across field strengths and methods. Both the vendor and MRQuantif methods showed a strong correlation for R2* between 1.5 T and 3 T. The regression slope for R2* measurements between 3 T and 1.5 T was close to 2 (1.72 for maps and 1.83 for MRQuantif). LIC values derived from R2* showed high reproducibility between 1.5 T and 3 T only when using the identical calibration group. For LIC calculated from R2* using the same calibration group (European calibrations), the bias between field strengths remained low (maximum ±7%). The MRQuantif method showed an excellent correlation (R=0.99) and the narrowest Limits of Agreement (LoA) ([-26%, 17%]) between 1.5 T and 3 T. A notable bias of 26% arose when comparing LIC values obtained from American and European calibrations. This discrepancy is significantly greater than the small differences observed across methods or field strengths. For clinical convenience, a rapid and intermediate LIC (in μmol/g) can be estimated by dividing R2* values by 2 at 1.5 T or by 4 at 3T. Limitations of the study included the lack of liver biopsy as reference standard. However, liver biopsies would not have been ethically justified as MRI-derived LIC and PDFF have been validated before and are widely accepted in clinical practice. Another limitation is the use of a first TE around 1 ms. A shorter first TE could have mitigated potential errors in high iron overload. Furthermore, MRI systems of only two vendors have been compared. In conclusion, LIC values derived from R2* are highly comparable between field strength and vendors when similar calibration methods (European or American) are applied. The use of the vendor-independent method (MRQuantif) provides the highest correlation for LIC, especially in patients with high iron overload. The significant bias between results from different calibration groups highlights a crucial issue. The conversion formula chosen must align with the referring physicians' standard references (American or European) to avoid disrupting patient care. The authors recommend that the R2* value and the name of the calibration formula used to calculate LIC should be clearly mentioned in the report. | References
|