Eur Radiol (2011) 21: 137–141 DOI 10.1007/s00330-010-1899-z MAGNETIC RESONANCE Agustin Castiella Jose M. Alústiza Jose I. Emparanza Eva Mª Zapata Belen Costero Maria I. Díez Received: 18 March 2010 Revised: 6 May 2010 Accepted: 2 June 2010 Published online: 6 August 2010 # European Society of Radiology 2010 Liver iron concentration quantification by MRI: are recommended protocols accurate enough for clinical practice? Abstract Objective To assess the accuracy of quantification of liver iron concentration (LIC) by MRI using the Rennes University (URe- nnes) algorithm. Methods In the overall study period 1999–2006 the LIC in 171patients was calculated with the URennes model and the results were compared with LIC measured by liver biopsy. Results The biopsy showed that 107 patients had no overload, 38 moderate overload and 26 high overload. The correlation between MRI and biopsy was r =0.86. MRI correctly classified 105 patients according to the various levels of LIC. Diagnostic accuracy was 61.4%, with a tendency to overestimate overload: 43% of patients with no overload were diagnosed as hav- ing overload, and 44.7% of patients with moderate overload were diag- nosed as having high overload. The sensitivity of the URennes method for high overload was 92.3%, and the specificity for the absence of overload was 57.0%. MRI values greater than 170 μmol Fe/g revealed a positive predictive value (PPV) for haemochromatosis of 100% (n =18); concentrations below 60μmolFe/g had a negative predictive value (NPV) of 100% for haemochromatosis (n =101). The diagnosis in 44 patients with inter- mediate values remained uncertain. Conclusions The assessment of LIC with the URennes method was useful in 74.3% of the patients to rule out or to diagnose high iron overload. The method has a tendency to overestimate overload, which limits its diagnostic performance. Keywords Haemochromatosis . Liver biopsy . Liver diseases . Magnetic resonance imaging . Comparative studies Introduction Quantification of liver iron concentration (LIC) with magnetic resonance imaging (MRI) is a non-invasive technique that was developed in the 1980s [1, 2]. Fundamental research into the development of the technique includes that conducted by the team of Rennes University, France, led by Gandon [3]. This team has designed an MRI technique for any 0.5-, 1-, or 1.5-T MR system, and has developed a free-access, Internet-based calculation model [4]. This recommended protocol is now widely used in clinical practice, and many studies use the results obtained with it as the reference value for LIC [5– 8]. It is considered reliable for LIC values under 350 μmol Fe/g; because of its simplicity and reproduci- bility, it is a useful option for standardising the quantifi- cation of LIC with MRI. Very few studies have compared the results of this model with those of the gold standard (LIC measured by biochemical analysis from a liver biopsy) [3, 9–12]. However, all diagnostic tests must be assessed more than once for reliability, reproducibility and accuracy with respect to the reference standard. The objective of this study was to assess the accuracy of LIC estimated by MRI with the Rennes University calculation model by comparing it with LIC measured A. Castiella : E. M. Zapata Gastroenterology Service, Mendaro Hospital, Mendaro, Spain J. M. Alústiza ()) Radiology Service, Osatek Donostia, Paseo Dr. Beguiristain, Donostia, Spain e-mail: jmalustiza@osatek.es Tel.: +34-94-3007075 Fax: +34-94-3007258 J. I. Emparanza Clinical Epidemiology Unit, Donostia Hospital CASPe, CIBER-ESP, Donostia, Spain B. Costero Gastroenterology Service, Principe de Asturias Hospital, Alcala de Henares, Spain M. I. Díez Radiology Service, Principe de Asturias Hospital, Alcala de Henares, Spain