young children, who would otherwise require sedation in order to remain still for the duration of a long T 2 -MRI examination. Our results demonstrate that T * 2 -measured LIC is as reliable as T 2 -derived measurements that are the current non-invasive gold-standard. Furthermore, our T * 2 method is highly robust, showing excellent reproducibility between operators and between repeated assessment by the same operator. Our results corroborate previous findings [4,9,10] and provide further evidence for the use of T * 2 -MRI for rapid, accurate, and reproducible non-invasive LIC quantification in iron-overloaded patients.
Methods
Patient population. Seventy-five patients, aged 12.4 ± 5.0 (range, ages 5-23), with iron overload were enrolled in this Institutional Review Boardapproved prospective study. The diverse population included patients with thalassemia major (38), thalassemia intermedia (3), sickle cell disease (23), Diamond-Blackfan anemia (3), Fanconi anemia (2), hereditary spherocytosis (1), autoimmue hemolytic anemia (1), congenital sideroblastic anemia (1), pyruvate kinase deficiency (1), Hodgkin lymphoma (1), and hereditary hemochromatosis (1). Of these patients, 70 received frequent red cell transfusions.
MRI technique. T 2 and T * 2 data were acquired on all patients on a clinical 1.5 Tesla Siemens scanner (Avanto). The T 2 protocol employed a multi-slice spin-echo sequence (scan time 5 30 min 20 sec): 11 axial slices, slice thickness (TH) 5 5 mm, matrix 5 192 3 256, repetition time (TR) 5 2500 msec, number of excitations (NEX) 5 1; repeated for five echo times (TE) 5 6, 9, 12, 15, 18 ms. The T * 2 protocol employed a non-breath-held multi-slice multiecho gradient-echo sequence (scan time 5 3 min 22 sec): 11 axial slices, TH 5 6 mm, matrix 5 101 3 192, variable field-of-view (FOV) (default 5 350 mm), flip angle 5 608, TR 5 500 msec, NEX 5 4; 11 equally spaced TE 5 (2.3-30) msec.
Reference LIC calculations based on T 2 data were provided by FerriScan 1 . T * 2 data were analyzed using in-house software (Matlab v.7.0) on a single axial slice chosen by the observer. Analysis was based on the optimal method described in Ref. 11 for achieving accuracy. R * 2 (1/T * 2 ) was computed at every pixel location using a constant offset model (S 5 S o e 2TE 3 R2* 1 C) (eq. 1). A region-of-interest (ROI) was manually drawn on the R * 2 map to encompass the entire liver, excluding obvious blood vessels and ducts. The LIC (mg/g) was computed from the ROI-median R * 2 value using the calibration curve [Fe] 5 0.0254 R * 2 1 0.202 (eq. 2) (Ref. 4). Two independent observers performed analysis and prescribed ROIs with no knowledge of FerriScan's results. One observer repeated the analysis in each patient on a different axial slice to determine slice-dependent and intra-observer variation.
Statistical analyses. Descriptive statistics were used to describe the sample. Inter-observer and intra-observer agreement of measurements was assessed in terms of the intraclass correlation coefficient (ICC). Pearson correlation was used to measure the agreement between T * 2 -measured and reference LIC. Linear regression was used to estimate the slope of the regression line fitted to the absolute LIC measured by the T * 2 method versus reference T 2 -based measurements.