Comparison of 3 T and 1.5 T for T2* magnetic resonance of tissue iron

A total of 104 consecutive subjects were prospectively recruited from referrals for siderosis screening. Furthermore, a group of 20 healthy volunteers was included. Patients were already scheduled to undergo T2* MR myocardial and liver content assessment at 1.5 T for clinical purposes and a research scan at 3 T was performed in addition. Inclusion criteria were: 1) age >18 years old and 2) written and informed consent to participate. Exclusion criteria were: 1) claustrophobia, 2) metallic implants or permanent pacemaker and 3) inability to hold recumbent position for >15 min. The study protocol was approved by North-East Thames ethics committee.

The T2* MR protocol consisted of two parts: 1) Myocardial and liver iron content assessment on a 1.5 T Sonata scanner 2) Repeated myocardial and liver iron burden assessment with a 3 T Skyra scanner (both Siemens Medical Systems, Erlangen, Germany).

The analysis of myocardial and liver T2* has been described elsewhere [4, 12]. In brief, dedicated software (Thalassaemia tools, a plug-in of CMRtools, Cardiovascular Imaging Solutions, London) was used. The entire thickness of the cardiac interventricular septum and a large region of interest within the liver parenchyma without blood vessels were selected to measure signal intensity (SI) at each echo time. The SI was plotted against echo time and a mono-exponential trendline was fitted to the decay curve to derive T2* according to this equation:

T2* was subsequently transformed into its reciprocal R2* according to the equation:

Curve fitting was done according to the truncation method [13]. In patients with cardiac T2* < 10 ms or liver T2* < 3.3 ms at 1.5 T, a second-moment noise-corrected model was used [14]. Cardiac T2* image quality was prospectively evaluated for each acquisition according to the following scoring scale: 0-unusable: uninterpretable images, 1-poor: heart just discernible, 2-average: severe septal artefact, 3-good: moderate septal artefact, 4-very good: mild septal artefact: 5-excellent: no septal artefact. Higher scores therefore indicated less artefact and higher image quality.

In order to compare reproducibility at 1.5 T and at 3 T, a group of 20 patients was selected for inter-observer, intra-observer variability and for inter-study variability. The 20 selected patients had levels of myocardial and liver siderosis that covered the entire range of the disease (from absent to severe iron overload). For intra-observer variability, the same investigator was asked to reassess myocardial iron content on WB and BB sequences and liver iron at 1.5 T and 3 T within a month after the first scan. For inter-observer variability two investigators experienced in MR iron content evaluation separately reported T2* at each sequence at 1.5 T and 3 T for each selected patient. Finally, patients underwent a repeated 3 T cardiac and liver iron content evaluation after a period of 1 h after the previous scan, for inter-study reproducibility assessment.

Imaging data were not normally distributed and are therefore expressed as median (interquartile range). Comparison of T2* and R2* values between normal volunteers and iron overload patients was performed with Mann-Whitney U-test. Association between T2* and R2* values according to different sequence acquisition and field strengths were evaluated with the Spearman’s correlation coefficient with 95 % confidence interval (CI) or non-linear regression when appropriate. Cardiac artefact scoring between different acquisitions and field strengths were compared using a Chi-Square test. Coefficients of variation (CoV) and intra-class correlation (ICC) (with 95 % CI) analysis were conducted to assess inter- and intra-observer variability in addition to inter-study variability of cardiac and liver T2* values at 1.5 T and 3 T. For ICC, an alpha two-way mixed model with absolute agreement analysis was used. Statistical comparison between BB sequence inter-and intra-observer and inter-study reproducibility at 1.5 T and at 3 T was performed using the squared difference between the 2 observations as an estimate of between subjects variance and was multiplied by 2. A paired T-Test was done to compare the different squared difference. When appropriate, the squared differences were log transformed to allow T-test use. When the difference was zero, it was replaced by half of the next value before log transformation [15]. Finally, Bland-Altman plots were generated to further describe the reproducibility of T2* measurements at 3 T [16]. A P-value <0.05 was considered significant. All analyses were done using SPSS software version 22.0.0, IBM, Chicago, Illinois, USA.

A total of 124 subjects were included in the analysis. Demographics, clinical characteristics and median heart and liver T2* measurement at 1.5 T and at 3 T in both iron overload (n = 104) and normal (n = 20) patients are shown in Table 1. β-thalassemia major was present in 45 %. In the patients with iron overload, median age was 30 (23–53) years old. The majority of patients (73 %) were treated with at least one iron chelating agent. There were 7 patients with severe cardiac iron overload (cardiac T2* < 10 ms at 1.5 T), and 45 patients with severe liver iron overload (liver T2* < 3.3 ms at 1.5 T). Liver acquisitions were not performed in 4 patients due to a temporary scanner fault not related to the research.

Using the WB sequence, the relationship between T2* values at 1.5 T and 3 T for the heart was non-linear with good fit (R ² = 0.954, P < 0.001) (Fig. 1a). There was good linear correlation between R2* values at 1.5 T and at 3 T with the WB sequence (R ² = 0.971, P < 0.001) (Fig. 1b) with near doubling (94 %) of R2* values at 1.5 T. Association between T2* and R2* values at 1.5 T and 3 T using the BB sequence are displayed in Fig. 2a and b. The relationship between BB T2* values at 1.5 T and 3 T was non-linear with good fit (R ² = 0.931, P < 0.001). Good linear correlation was found for BB R2* measurements at 1.5 T and at 3 T (R ² = 0.979, P < 0.001) again with near doubling (94 %) of R2* values at 1.5 T.

Regression graphs between T2* and R2* at 1.5 T and 3 T for the liver are displayed in Fig. 3a and b. For T2*, an excellent non-linear fit was found between the values obtained at 1.5 T and at 3 T (R ² = 0.993, P < 0.001). Similar to heart R2* values, excellent linear correlation was found between R2* values at the two different field strengths (R ² = 0.993, P < 0.001) with an increase of 105 % in R2* from 1.5 T to 3 T.

Median artifact score was 4 (4–5) with the 1.5 T BB sequence, 4 (3–4) with the 3 T WB sequence and 4 (4–5) with the 3 T BB sequence. The artifact score was significantly higher (higher indicates less artefact) with the BB sequence at 1.5 T than with the WB sequence at 3 T (P = 0.025) and the BB sequence at 3 T (P = 0.007). Moreover, the artefact score was significantly superior with the BB sequence at 3 T than with the WB sequence at 3 T (P < 0.001) (Fig. 4).

The evaluation of T2* intra-observer variability is shown in Table 2. There was excellent agreement between the repeated measurements in all sequences at both 1.5 and 3 T, with mildly increased CoV (inferior reproducibility) at 3 T compared to 1.5 T. The Bland-Altman plots confirm that there was no systemic bias in repeated assessment of T2* at 3 T by the same observer (Figs. 5a, 6a, 7a).

Table 3 displays the CoV and ICC for estimation of T2* inter-observer variability. There was excellent agreement between the two observers. Bland-Altman plots confirms the agreement between the observers at 3 T (Figs. 5b, 6b and 7b) without the presence of systematic bias. Finally, inter-study reproducibility was excellent as shown in Table 4. CoV was found to be slightly inferior at 3 T than at 1.5 T. Overall, at 3 T, the BB imaging sequence had less variability of cardiac T2* measurements than the WB sequence but this did not achieve statistical significance (Table 5).

The present study was conducted in a single centre with experience in T2* assessment. The relatively low CoV and high ICC values obtained should ideally be compared to the reproducibility data from other centres. One advantage conferred by 3 T MR may not have been fully explored in the present study, which is the possible increase of the acceleration factor and consequent shortening of the acquisition time. Such an increase in image acquisition speed however, may lead to reduction in signal to noise ratio and loss of image quality. Only adults >18 years were studied in this research which may limit the conclusions that may be drawn in younger and smaller patients.