Key words MR-imaging - iron - abdomen - hematologic
Introduction
Noninvasive quantification of liver iron content (LIC) is essential for patients with iron overload due to increased iron resorption or patients suffering, e. g., from anemia receiving blood transfusions. MRI methods are widely used for noninvasive LIC quantification. Overviews are given, e. g., by Sarigianni and Henninger [1 ]
[2 ]. The majority of them use gradient-echo sequences, while a method employing spin-echo has been FDA-certified [3 ].
As far as we figured out, the refocused gradient echo sequence, also called balanced steady-state free precession (bSSFP) sequence, has not yet been investigated for its potential to quantify iron. This sequence is characterized by a high SNR efficiency, i. e., a high ratio of signal-to-image noise per scan time [4 ]. For this reason, the suitability of the bSSFP sequence for LIC quantification was investigated in this study.
Methods
According to the latest version of the Declaration of Helsinki (revision dated 2013, Fortaleza, Brazil) and after approval of our local ethics committee, patients referred to our institution for noninvasive LIC quantification between June 2014 and February 2015 were studied at 1.5 T (Magnetom Avanto, Siemens Healthcare GmbH, Erlangen, Germany). All patients, or, for minors, their parents, gave written informed consent. LIC was determined using the FerriScan procedure [3 ]. Images were acquired with the appropriate protocols and transferred for centralized evaluation. These LIC values served as a reference. Additionally, eleven protocols with different contrasts were acquired in all patients with a bSSFP sequence in 3D mode with FoV of 315*420 mm, matrix size 192 × 112 × 26 and resolution 2.2 × 2.2 × 5 mm. Acquisition was performed in shallow breathing with long-term averaging of 3 acquisitions to avoid pulsation and breathing artifacts, resulting in a total acquisition time of seven minutes for all protocols together. Parameters deviating between the different bSSFP protocols are listed in [Table 1 ]. For signal homogeneity in the imaging plane, the body resonator was used as a receive coil.
Table 1
List of bSSFP protocols used.
Protocol no.
Echo time TE [ms]
Repetition time TR [ms]
Bandwidth [Hz/pixel]
Acquisition time [s]
Excitation flip angles FA [°]
1–3
1.75
3.5
668
29
7, 10, 17
4, 5
1.75
3.5
1184
29
30, 50
6–8
2.5
5
395
41
7, 17, 60
9–11
3.25
6.5
243
53
7, 17, 50
Overview of the bSSFP protocols used in this study.
The bSSFP data were retrospectively analyzed by manually positioning three circular regions of interest (ROIs) of predefined size (1.75 cm2 ) in the vessel-free liver parenchyma, preferably the right liver lobe, and two bilaterally in the paraspinal musculature (0.75 cm2 ), as shown in [Fig. 1 ]. ROI placement was performed in two different axial slices in each patient. Subsequently, the liver signal was divided by that of the muscle to obtain signal intensity ratio (SIR) values.
Fig. 1 Example of a bSSFP image at a TR of 3.5 ms and 17° FA with three liver ROIs and two in the paraspinal muscles.Abb. 1 Beispiel eines bSSFP-Bildes mit einer Repetitionszeit TR von 3,5 ms und einem Anregungswinkel von 17° mit 3 ROIs in der Leber und 2 in der paraspinalen Muskulatur.
Statistical analysis was performed in SPSS (v. 27, 2020, IBM, Armonk, USA). Logarithms of SIR values of the different protocols were linearly correlated with the liver iron reference value. The coefficient of determination R2 was used to determine the quality of correlations. Protocols showing the best correlation at this step were then used to test multiple linear correlations of SIR logarithms to LIC reference. Correlation parameters of this multiple linear correlation were used to derive a relationship to determine the LIC from bSSFP data. The sensitivity and specificity of these bSSFP-LIC values with respect to the reference method were studied for an LIC threshold of 80 µmol/g (4.5 mg/g) relevant for iron chelation therapy [5 ].
Results
In the studied time interval, data of 35 patients (24 male, age range: 6.9 to 73 years, mean ± standard deviation 24.2 ± 14.4 years) were suitable for analysis and were evaluated.
LIC reference values ranged from 24 to 756 µmol/g (MW ± stdDev 145 ± 143 µmol/g) in our patient cohort. Yet, their distribution was not uniform: most LIC reference values were observed between 20 and 200 µmol/g, only a few were between 200 and 400 µmol/g, and there was only one rather high value of 756 µmol/g.
The best correlation of a single protocol was obtained with a repetition time TR of 3.5 ms at a flip angle (FA) of 17° with a coefficient of determination R2 of 0.82. [Fig. 2 ] shows a scatterplot of these data. For the other TRs, FA of 17° also yielded the best results. A multiple linear correlation of the three protocols with TRs of 3.5, 5, and 6.5 ms, each at an FA of 17°, resulted in a coefficient of determination of 0.85. [Fig. 3 ] shows a scatterplot of bSSFP-LIC values determined using the resulting relationship, and the corresponding 2-by-2 table for the LIC threshold of 80 µmol/g (4.5 mg/g) is given in [Table 2 ]. We found a sensitivity/specificity of 0.91/0.85.
Fig. 2 Scatterplot of LIC reference values vs. ln (SIR). The dotted line is the regression line.Abb. 2 Punktegrafik der Referenzwerte der Lebereisenkonzentration als Funktion der Logarithmen der Signal-Intensitäts-Verhältnisse. Die gepunktete Linie ist die Regressionsgerade.
Fig. 3 Scatterplot of LIC reference values vs. bSSFP-LIC. The dotted line is the regression line.Abb. 3 Punktegrafik der Referenzwerte der Leber-Eisen-Konzentration gegenüber den mit der bSSFP-Methode ermittelten Werten. Die gepunktete Linie ist die Regressionsgerade.
Table 2
2-by-2 table: bSSFP vs. reference method.
LIC bSSFP
Negative
Positive
Sum
LIC reference
Neg
11
2
13
Pos
2
20
22
Sum
13
22
35
2 × 2 table of LIC values calculated from bSSFP data vs. LIC reference values for the threshold of 80 µmol/g (4.5 mg/g), important for therapy management.
Discussion
The balanced gradient echo sequence (bSSFP) has a complex signal behavior. Due to the necessary balance of all gradients, it is favorable to choose the echo time TE as half the repetition time TR. In this case, the bSSFP signal does not depend on T2 *, but on T2 , like spin-echo. However, contrast is influenced also by longitudinal relaxation T1 . In detail, contrast depends on the T2 to T1 ratio [6 ]
[7 ]. Another parameter influencing contrast is the excitation angle, which is freely selectable within the maximum possible energy deposition. Short echo times are advantageous for this sequence, but require fast gradient switching and imply high energy exposure to the patients because of the fast repetition rate of excitation pulses. Hence, the maximal flip angle is limited.
As the presented data show, bSSFP is in principle suitable to quantify liver iron content. Its advantages are high scanning efficiency, i. e., high signal-to-noise ratio achieved in short acquisition times. This allows imaging of the entire liver in a single breath-hold without additional acceleration techniques such as parallel imaging, which require multiple receive coils and multiple receiver channels, but increase image noise. In our approach, breath hold could not be used in order to minimize pulsation artifacts. This was efficient for short TR protocols, but worse with increasing TR (data not shown).
The sequence used was a standard sequence that is generally available and easily applicable in the clinical routine.
In total, eleven protocols were studied since it was not clear which combination of acquisition parameters was best suited for LIC quantification. Originally, it was intended to determine the tissue parameters T1 and T2 with a nonlinear fitting process. To evaluate the feasibility of this, multiple parameter combinations seemed necessary and advantageous. Unfortunately, we were not successful (data not shown). Probably, this was caused by RF inhomogeneity, which was present in some patients at varying locations, and/or by banding artifacts. For SIR analysis, which also was planned from the beginning, it was unclear which parameters perform best and if, for the wide range of LIC values expected for our patients, a step-wise approach as proposed by Gandon would have been required [8 ].
The paraspinal muscle signal as internal reference was proposed by Alustiza and Gandon to quantify LIC with GRE [8 ]
[9 ]. The motivation to investigate the correlation of the natural logarithm of SIR values rather than the SIR values themselves was previously published work on the SIR of gradient echo data [10 ].
The sensitivity and specificity previously reported for a sophisticated spoiled gradient echo (GRE) method requiring only a single breath hold were 0.98/0.67 for the 80 µmol/g (4.5 mg/g) threshold [11 ]. In this preliminary study we found superior specificity for bSSFP, which, however, still needs to be verified in larger patient numbers. Sensitivity was higher for GRE. Probably, a combination of GRE and bSSFP would be advantageous for reliable patient management.
The major advantage of the method presented here is the simple mathematical approach avoiding an extensive fitting process. Compared to other SIR approaches at 1.5 T, bSSFP has the benefit of covering a large LIC range, avoiding the need for an additional protocol to address excessive iron overload [12 ]. With these benefits, bSSFP is probably suited for liver iron overload screening.
Limitations
The high sensitivity of the bSSFP sequence to magnetic field inhomogeneities is probably one reason for the scatter of the measured SIR values, since a standard bSSFP sequence was used. Maybe, results could be enhanced using proposed measures to reduce bSSFP signal reduction caused by susceptibility effects [13 ]
[14 ].
The number of patients studied was small. Furthermore, LIC reference values were not normally distributed. In detail, a single value was nearly double the second largest. Therefore, the obtained calibration equation may be not very reliable. Further studies with larger patient numbers and a distribution of LIC reference values closer to Gaussian distribution are needed to evaluate the suitability of the bSSFP sequence more thoroughly.
Despite these limitations, we were able to show that bSSFP, an efficient and fast sequence that is, however, sensitive to susceptibility effects, is suitable for the quantification of liver iron content. Sequence variants which are more robust against susceptibility artifacts should be studied for their potential to improve the method.
