Purpose

This work was performed to test the influence of acquisition parameters on a) image quality of gradient echo (GRE) acquisitions and b) the correlation between liver-to-reference signal intensity ratio (SIR) values obtained with GRE and reference liver iron concentration (LIC) based on spin echo (SE). Correlation was then c) also tested for influence of image quality.

Background

It has been shown previously that there is a linear correlation between natural logarithm of SIR values (ln SIR) and LIC (1). In brief, slope of ln (SIR) vs. LIC is associated to R₂* differences of tissues, whereas intercept of this relation is the logarithm of scaling factor (S₀) ratio, which depends on tissue T₁ and MR acquisition parameters.

Methods

85 patients (38 f, 47 m, age range 2.3 – 73 years, mean 24 ± 15.7 y) suspected for liver iron overload were scanned at 1.5 T (Siemens Avanto, Siemens Healthineers, Iselin, NJ) with two MRI protocols: a) the Ferriscan® protocol involving five SE sequences in free breathing, each of 3:15 min duration, and b) three transversal slices acquired using multi-echo GRE at TE/TR 4.7-9.5-14.2-19/120 ms, FA of 20° and 90°, both with and without RF spoiling (FLASH/T1-FFE/SPGR vs. FISP/FFE/GRASS), requiring a total of four breathholds. A saline bag was positioned on the left side adjacent to the patient body. In 31 investigations, saturation regions (SATs) were applied cranial and caudal of the transversal slices in order to destroy signal from mobile protons causing pulsation artefacts. Image quality was scored on a five-point scale (no/minor/moderate/severe artifacts, non-diagnostic). Three regions were manually placed in the liver, one each in left and right paraspinal muscle, and one in the saline bag. ROI median value was employed to calculate natural logarithm of liver-to-muscle and liver-to-saline SIR. Both were correlated to reference LIC obtained by Ferriscan®. Linear regression analysis of ln (SIR) vs. reference LIC was performed separately for different FA, RF spoiling condition, +/- SATs, and image quality score ranges, yielding values for slope and intercept of linear regression lines and their uncertainties, as well as coefficient of determination R². Results of regression were compared between acquisition parameters and image quality scores. Differences, where present, were checked for significance.

Results

With RF spoiling, image quality was scored 1 or 2 (no or minor artefacts) in 23/31 (74%) investigations performed with SATs and 25/54 (46%) scans without SATs. Overall, 48/85 (56%) examinations were scored 1 or 2. Without RF spoiling, only 19/31 (61%) cases with SATs, and 22/54 (41%) without SATs, overall 41/85 (48%) were scored 1 or 2. No case was scored non-diagnostic, neither with nor without RF spoiling. For mean and SD of scores for different conditions cf. Tab. 1. Slopes of ln (SIR) vs. reference LIC were equal between conditions, whereas intercept varied between acquisitions with and without RF spoiler at FA 90°. At FA 20°, there was no significant difference between RF spoiling conditions. When using saline reference, differences concerning RF spoiling at FA 90° were more pronounced than using paraspinal muscles as reference. Some examples of correlations between ln (SIR) and reference LIC are depicted in Fig. 1 for liver-to-NaCl SIR and in Fig. 2 for liver-to-muscle SIR. Correlations between reference LIC and ln (SIR) were good, with an R² between 0.85 and 0.9 with saline reference, and 0.88 < R² < 0.93 with paraspinal muscles as reference. Within their uncertainty, correlation parameters slope and intercept, as well as R², were independent of image quality score, +/- SATs, and +/- RF spoiling. Uncertainty of correlation parameters was lower with muscle reference than saline.

Discussion

Correlation of signal intensity ratios determined from GRE MRI to reference LIC values determined with SE turned out to be independent of GRE scan parameters. Although saturation regions and RF spoiling have a positive impact on image quality, it was shown that they do not influence correlation between ln (SIR) and reference LIC. Coefficients of determination and precision of correlation parameters benefit from using paraspinal muscles as signal reference compared to a saline bag.

Acknowledgements

No acknowledgement found.

References

1. Wunderlich AP, Cario H, Bommer M, Beer M, Schmidt SA, Juchems MS. MRI-Based Liver Iron Content Determination at 3T in Regularly Transfused Patients by Signal Intensity Ratio Using an Alternative Analysis Approach Based on R₂* Theory. RoFo 2016. PubMed PMID: 27299667.