To evaluate the effectiveness of the single ROI approach for the detection of hepatic iron burden in thalassemia major (TM) patients in respect to a whole liver measurement.
To evaluate the effectiveness of the single ROI approach for the detection of hepatic iron burden in thalassemia major (TM) patients in respect to a whole liver measurement.
Five transverse hepatic slices were acquired by a T2* gradient-echo sequence in 101 TM patients and 20 healthy subjects. The T2* value was calculated in a single region of interest (ROI) defined in the medium-hepatic slice. Moreover, the T2* value was extracted on each of the eight ROIs defined in the functionally independent segments. The mean hepatic T2* value was calculated.
For patients, the mean T2* values over segments VII and VIII were significantly lower. This pattern was substantially preserved in the two groups identified considering the T2* normal cutoff. All segmental T2* values were correlated with the single ROI T2* value. After the application of a correction map based on T2* fluctuations in the healthy subjects, no significant differences were found in the segmental T2* values.
Hepatic T2* variations are low and due to artifacts and measurement variability. The single ROI approach can be adopted in the clinical arena, taking care to avoid the susceptibility artifacts, occurring mainly in segments VII and VIII. J. Magn. Reson. Imaging 2011;33:348–355. © 2011 Wiley-Liss, Inc.
PEOPLE WITH THALASSEMIA major (TM) require regular blood transfusions, which can lead to iron overload in tissues because humans have no mechanism for eliminating excess iron (1). Excess body iron is highly toxic (2). This toxicity involves many organs such as the liver and heart, leading to a variety of serious diseases, which can be managed effectively or prevented. Therefore, early diagnosis of iron overload and appropriate and tailored chelation therapy are critical (3).
MRI is the only technique that quantitatively and noninvasively assesses both myocardial and liver burden (4). MRI multi-echo T2-star (T2*) sequences allow rapid monitoring of iron deposition, exploiting the fact that paramagnetic iron compounds produce variability in susceptibility, shortening the T2* relaxation time (5). For the heart, a multislice approach was set up (6, 7) to detect the heterogeneous myocardial iron distribution shown by histological studies (8, 9). For the liver, a single transverse slice is traditionally acquired in clinical practice and the T2* measurement is performed in a region of interest (ROI) of the parenchyma (7, 10, 11). The use of the conventional single ROI technique has been partially due the fact that the early development of the T2* sequence in single echo multi-breathhold required approximately nine 10-s breathholds to acquire a single slice. Maris et al averaged T2* values obtained from two ROIs placed in two orthogonal slices (12). The T2* obtained was taken as representative of the T2* value for the whole liver and has been shown to correlate with iron overload assessment performed through biopsies (10, 11).
Liver iron overload estimation by the ROI-based approach may suffer from sampling errors due to the choice of slice location and the user-dependent placement of the ROI. In particular, heterogeneity in liver iron deposition could be missed in single ROI measurements. Several histological studies have reported an intra-organ variability in HIC (hepatic iron concentration) measurements performed by biopsies in cirrhotic livers (13–15). Ambu et al also found uneven iron distribution in the liver of two beta-thalassemic patients without cirrhosis (16). Anyway, heterogeneity among the T2* values in the liver could be generated not only by uneven iron distribution, but also by additive susceptibility artifacts.
On the basis of these observations, we set-up a multislice approach characterized by the acquisition of several hepatic slices using a multi-echo single breathhold sequence in a large population of TM patients. This approach allowed detection of the hepatic segments, and the measurement of the T2* value in each of them. Thus, the aim of this study was to compare the effectiveness of the single ROI technique used in the clinical practice with the developed segmental approach.
One hundred one thalassemia major patients (48 males and 53 females, age 10–49 years, mean age 29 ± 8 years) were studied. All patients were enrolled in the MIOT (Myocardial Iron Overload in Thalassemia) network, constituted by thalassemia and MRI centers where MRI exams are performed using homogeneous, standardized and validated procedures and where patients' clinical-instrumental data are collected in a centralized database by means of the Web (17). All TM patients had been regularly transfused since early childhood and began chelation therapy from the mid-to-late 1970s, while patients born after the 1970s received chelation therapy from early childhood. Overall duration of chelation treatment was 23.6 ± 10.5 years. Over the past year, the mean serum ferritin level was 1413 ± 1209 ng/mL and the mean pretransfusion hemoglobin level was 9 ± 1 g/dL. Twenty-seven TM patients (27%) suffered from chronic hepatitis. In addition, 20 healthy volunteers (12 males; mean age, 28 ± 7 years) were involved in the study.
The study complied with the Declaration of Helsinki. All subjects gave written informed consent to the protocol. The project was approved by the institutional ethics committee.
MRI acquisitions were performed at five sites (Pisa, Catania, Ancona, Campobasso, and Palermo) included in the MIOT network. Inter-site reproducibility within the network was previously assessed (18). MRI was performed using a 1.5 Tesla (T) MRI scanner (GE Signa/Excite HD, Milwaukee, WI). An eight-element cardiac phased-array receiver surface coil with breathholding in end-expiration was used for signal reception.
Five transverse slices through the liver were obtained by a T2* gradient-echo multiecho sequence (Fig. 1). The five MRI sites used a research sequence provided by GE Healthcare within the MIOT study. A commercial version of this sequence is now commercially available. Each single slice was acquired at different echo times (TEs 2.0–21 ms with an echo spacing of 2.26 ms) in a single end-expiratory breathhold. The echo spacing was chosen equal to the separation between in-phase and opposite-phase condition of fat/water interface to minimize the shift of signal decay. The multi-echo sequence parameters were as follows: flip angle 25°, matrix 192 × 192 pixels, field of view (FOV) 40 × 40 cm, bandwidth 62.5 KHz, slice thickness 8.0 mm, slice spacing between 15 and 20 mm, depending on the liver dimension, number of excitations (NEX) 1, repetition time 25 ms (19). Mean acquisition time was approximately 3 min. T2* measurement was performed with a custom-written and previously validated software program (Hippo-MIOT IFC-CNR®; 7,19). The software was able to extract a decay curve by averaging the signal intensity at various TEs within a ROI of standard dimension manually positioned by the operator. The calculated decay curve was fit to a single exponential with a constant offset model:
where S represents the mean signal intensity, S0 is the estimated signal intensity at TE = 0, T2* is the relaxation time, TE represents the echo times and C is a constant value that takes into account the rectification of MRI noise. Decay curve fitting was performed by the Levenberg-Marquadt method and the T2* value for the ROI was obtained (19, 20).
First, as common in clinical practice, a single ROI was defined in the medium-hepatic slice. Hereafter, the calculated T2* value is called single ROI T2*.
In the multislice analysis, one ROI was defined in each of the eight functionally independent segments in which the liver can be divided, according to Couinaud (Fig. 2) (21). The mean hepatic T2* value was obtained by averaging all segmental values.
The lower limit of normal for the single ROI T2* was fixed at 15.8 ms. This cut-off was evaluated on 30 healthy subjects during the validation of the MIOT network (18).
To assess intra- and inter-observer variability, data related to 25 patients were selected from the entire data set in the following way. The patient population was divided into five groups, based on the level of iron overload: severe (single ROI T2* < 1 ms), mild (single ROI T2* between 1 and 2 ms), moderate (single ROI T2* between 2 and 5 ms), acceptable (single ROI T2* between 5 and 15.8 ms), and no iron (single ROI T2* ≥15.8 ms). Five patients were randomly selected from each group to cover the entire clinical range of the T2* values. The 25 data sets were presented in random order to two expert operators and blindly analyzed to evaluate the inter-observer variability. To evaluate the intra-observer variability, one of the two operators re-analyzed the images 15 days later without knowledge of the results of the previous analysis.
Segmental T2* fluctuations in the liver could be explained by three main sources: true heterogeneity in liver iron deposition, effect of susceptibility artifacts, and measurement errors. We developed a susceptibility artifacts map to take into account artifactual T2* fluctuations. We adopted the same approach used to build a T2* correction map applied to cardiac images (22). Briefly, it could be expected that systematic drift of T2* segmental values from the mean T2* liver values in subjects without iron overload is due only to susceptibility artifacts. Hence, the map was built in the R2* (1/T2*) domain by using the data of the healthy subjects involved in the study. The segmental correction factor for the segment k (Δ〈R2〉) was defined as the sum of the averaged deviations of the measured R2* in the segment with respect to the mean R2* value:
where N represented the number of subjects, R2 was the R2* value in the segment k for the j-th subject, and R2 was the mean value measured in this subject.
Susceptibility artifacts are additive in the R2* domain, so the developed map can be used to compensate the systematic drift of T2* values due to the effect of susceptibility artifacts in patients with iron overload. In particular, corrected T2* values can be evaluated by transforming T2* values in the corresponding R2* ones, correcting the R2* values by subtracting the corresponding correction map values, and returning to T2* notation.
All data were analyzed using the SPSS version 13.0 statistical package. Continuous variables were described as mean ± standard deviation (SD).
In the reproducibility analysis, for each segment, as well as for the mean value, the variability between two different analyses was evaluated by calculating the coefficient of variation (CoV) and the interclass correlation coefficient (ICC). The CoV was calculated as the standard deviation of the absolute percentage differences between the two T2* values divided by their mean. The ICC was obtained from a two-way random effects model with measures of absolute agreement. An ICC ≥0.75 was considered excellent, between 0.40 and 0.75 good, and <0.40 unsatisfactory.
One-way repeated measures analysis of variance (ANOVA) was used to evaluate whether there was a significant difference between T2* values in different hepatic segments. First, Mauchly's test was used to test assumption of sphericity. When the significance level of Mauchly's test was <0.05 and the sphericity could not be assumed, Greenhouse-Geisser corrected results were taken. The Bonferroni adjustment was used in all pairwise comparisons. The percentage of deviation of the segmental T2* values from the mean hepatic T2* value was assessed for each subject as the ratio of the difference between the segmental and the global values to the mean value, multiplied by 100. The coefficient of variation (CoV) was extracted as indicated above. Correlation analysis was performed using Spearman's test. Comparisons between the groups were made by an independent-sample t-test. The Wilcoxon rank sum test was applied for continuous values with a non-normal distribution. χ2 testing was performed for noncontinuous variables. In all tests a two-tailed P <0.05 was considered statistically significant.
In all subjects, it was possible to draw a ROI in each hepatic segment. The results of the intra- and inter-observer variability analysis performed in the selected patients are summarized in Table 1. CoV for intra-observer variability ranged from 4% to 11.2% and CoV for inter-observer variability ranged from 6.9% to 17.6%. An excellent ICC was always obtained (>0.95).
|Intra-observer reproducibility||Inter-observer reproducibility|
|ICC||CoV (%)||ICC||CoV (%)|
The mean T2* segmental values evaluated over the whole patient population ranged from 7.8 ms (segment VII) to 10.5 ms (segment I) (Table 2). The ANOVA test showed significant differences in the segmental T2* values (P < 0.0001). Specifically, the mean T2* values over the segments VII and VIII were significantly lower than the mean T2* values over the other segments (Fig. 3a). The mean hepatic T2* value was 9.6 ± 9.6 ms. The percentage of deviation of the segments from the mean hepatic T2* ranged from −10.8% (segment VII) to 1.2% (segment IV). Figure 3b shows the random fluctuation in hepatic T2* among the segments. A strong and significant correlation was found among all the segmental and the mean T2* values. Table 2 shows the calculated coefficient correlations, the correspondent P values and the percentage of deviation of the segments from the mean hepatic T2*.
|Segment||T2* (ms)||Comparison with the mean hepatic T2* value||Comparison with the mean single ROI value|
|Mean||SD||Deviation (%)||r||P||Deviation (%)||r||P|
The mean single ROI T2* value was 8.6 ± 8.6 ms. The percentage of deviation of the segments from the single ROI T2* ranged from −3.0% (segment VII) to 17.5% (segment I). All mean segmental T2* values as well as the mean hepatic T2* value were strongly correlated with the single ROI T2* value (Table 2).
The mean hepatic T2* values as well as the single ROI T2* values showed a significant negative correlation with the serum ferritin levels (r = −0.733, P < 0.0001 and r = −0.731, P < 0.0001, respectively).
The CoV of all patients (diamond-shaped markers) was plotted versus the mean T2* values (Fig. 4).
Patient population was divided into two groups. Group A included 81 patients with abnormal single ROI T2* (<15.8 ms) and Group B included 20 patients with normal single ROI T2* (≥15.8 ms). The two groups were not different regarding sex (group A: M/F 40/41 versus group B: M/F 8/12, P = 0.618). Group A was younger than Group B (28 ± 7 years versus 37 ± 7 years, P < 0.001) and showed higher serum ferritin levels (1664 ± 1248 ng/mL versus 526 ± 361 ng/mL, P < 0.0001). In both groups the single ROI T2* values were significantly correlated with the serum ferritin levels (Group A: r = −0.646, P < 0.0001 and Group B: r = −0.600, P < 0.0001).
In Group A, the mean lower T2* value was detected in the segment VII, followed by the segment VIII. In Group B, the mean lower T2* value was detected in the segment VIII, followed by the segment VII. The order of the other segmental T2* values was the same in both groups (segments IV, III, V, VI, I, II).
A significant segmental variability was found in both groups (P < 0.0001). Specifically, in Group A (as in the whole patient population) the mean T2* values over segments VII and VIII were significantly lower than the mean T2* values over the other segments (Fig. 5a). In Group B, the mean T2* values over segments VII and VIII were significantly lower than the mean T2* values over the other segments, except segment IV (Fig. 5b).
In Group A, the mean hepatic T2* value was 5.4 ± 4.2 ms. The percentage of deviation of the segments from the mean hepatic T2* ranged from −7.9% (segment VII) to 4.0% (segment III). The mean single ROI T2* value was 5.2 ± 4.0 ms. The percentage of deviation of the segments from the single ROI T2* ranged from −2.2% (segment VII) to 11% (segment I). In Group B, the mean hepatic T2* value was 26.5 ± 4.8 ms. The percentage of deviation of the segments from the mean hepatic T2* ranged from −23.3% (segment VII) to 19.4% (segment II). The mean single ROI T2* value was 23.4 ± 7.6 ms. The percentage of deviation of the segments from the single ROI T2* ranged from −12.2% (segment VII) to 37.3% (segment II).
The absolute percentage of deviation of the mean hepatic from the single ROI T2* was significantly lower in Group A than in Group B (10.6 ± 8.7% versus 28.1 ± 28.6%, P = 0.017).
The mean T2* segmental values evaluated over the healthy subjects ranged from 19.3 ms (segment VIII) to 29.9 ms (segment I; Table 3). The ANOVA test showed significant differences in the segmental T2* values (P < 0.0001). Specifically, the mean T2* value over segment VII was significantly lower than the mean T2* values over the other segments, except segment IV, and the mean T2* value over segment VIII was significantly lower than the mean T2* values over segments I, II, V, and VI (Fig. 6a).
|Segment||T2* (ms)||R2* (Hz)||Correction factor (Hz)|
The mean hepatic T2* value was 24.9 ± 2.6 ms. The percentage of deviation of the segments from the mean hepatic T2* ranged from −22.9% (segment VIII) to 19.8% (segment I).
Table 3 shows the segmental correction factors calculated according to Eq. . As expected, no significant variability was detected across corrected segmental T2* values in healthy subjects (P = 0.780) (Fig. 6b). After the application of the correction map, the ANOVA test showed no significant differences in the segmental T2* values (P = 0.476) for previously defined group B used as validation dataset being composed by patients with normal single ROI T2*. Also the whole patient population (P = 0.251), the previously defined Group A (P = 0.073) as well as the Group A patients (N = 23) with chronic hepatitis (P = 0.111) showed no significant differences in the segmental T2* values after the application of the correction map.
Precise and effective measurements of iron overload in the liver, where iron deposition seems to be primarily noticeable (20), are important for the early diagnosis, treatment, and follow-up of thalassemia patients.
Liver iron concentration measured in a small part of a needle biopsy is currently considered the gold standard for this purpose and is taken to be representative of the mean iron concentration in the whole liver (23). However, different studies, involving multiple biopsies, have shown a heterogeneity in liver iron deposition in cirrhosis (14) and in hemochromatosis (16, 24). Thus, iron content determined in a small liver sample could not absolutely represent the mean hepatic iron concentration. MRI multi-echo T2* allows assessment of liver burden quantitatively and noninvasively. In clinical practice the T2* measurement is performed in a ROI of standard size, manually drawn in a homogeneous area of parenchyma without blood vessels (7, 11). As with the single liver biopsy, the ROI-based method may suffer from sampling errors. Positano et al evaluated dependence on the ROI placement in the T2* measurements on 40 thalassemia major patients, showing that the variability shown by the ROI-based approach is acceptable for use in clinical practice (19). However, this study was limited to a single transverse slice.
Thus, the goals of our study were to set up a MRI acquisition technique for the detection of iron burden in the whole liver by a multislice approach, to detect potential preferential patterns of iron deposits in the hepatic segments (21) and to compare the effectiveness of the single ROI technique used in the clinical practice with the developed segmental approach.
The proposed approach was feasible in all patients and in all healthy subjects involved in this study. Moreover, the intra- and inter-operator reproducibility found for this approach could be considered satisfactory (ICC > 0.95) for clinical purposes (Table 1). As expected the coefficient of variation for the mean T2* was smallest of all due to a “compensation” of outliers that may lead to high variability in a single ROI measurement.
A significant segmental variability was found in the whole patient population, with significant lower T2* values in segments VII and VIII (Fig. 3). The prevalence (80%) of patients with an abnormal single ROI T2* value (Group A) confirmed that hepatic iron overload is a very common finding in thalassemia. Group A patients had significant higher serum ferritin levels compared with the patients with normal single ROI T2* value (Group B). In fact, as widely reported in the literature, liver T2* is strongly and inversely correlated with serum ferritin concentration (25, 26). Moreover, Group A was younger and, consequently, had undergone chelation therapy for a shorter time.
In both groups, a significant segmental variability was found with similar patterns (see the third subsection Segmental Analysis in the Whole Patient Population of the Results section). The detected T2* variations may represent true heterogeneous iron density or may be explained by taking into account T2* measurement errors and susceptibility artifacts. In fact, T2* measurement errors may produce different segmental T2* values even for subjects with exactly the same “true” T2* value in all segments. As shown in Figure 4, the segmental CoV dependence from T2* resembles the expected variations due to measurement errors, as assessed in previous studies for single-ROI measurements (19). In fact, the acquisition technique commonly used in clinical practice is most accurate for measurements in the clinical range of T2* values due to technical constraints (19, 27, 28). Specifically, the relative high CoV at very low T2* values (<4 ms) may be explained by the lack of precision in T2* assessment at these T2* levels due to the inadequate minimum TE. The CoV is also expected to increase slightly in borderline patients or in patients without iron overload, because the signal decay is only partially sampled due the too-low maximum TE. However, the increase in the CoV at high T2* values can be more likely explained by susceptibility artifacts. In fact, T2* measurements are sensitive to susceptibility artifacts that might originate from several sources such as air-tissue interfaces. Segments VII and VIII are close to the right lung base and with the air contained inside, suggesting that the significant T2* reduction in these segments is probably artifactual. The fact that the T2* values in segment IV were not different from the T2* values in segment VII and VIII only in the patient group without hepatic iron overload could be explained by the fact the iron mask the artifact effect in the other group. Some artifact sources could affect segment IV, but they were not so strong to be significant in patients with iron overload.
To better explore this aspect, 20 healthy subjects were involved in our study. For them as well, a significant segmental variability was found, with lower T2* values in segments VII and VIII, the same ones detected within the patient population. Healthy subjects' data were used to build a segmental correction map to compensate for the artifact effects (see the last subsection Correction Map of the Results section). After the application of the correction map, no significant differences in compensated T2* values across the segments were detected in either healthy subjects, as expected, or patients of Group B (thalassemia major patients with normal single ROI T2*) used as validation dataset. The vanishing of segmental variations confirmed its artifactual nature. The segmental variability disappeared also for subjects in Group A (thalassemia major patients with abnormal single ROI T2*) and even in Group A patients with chronic hepatitis where a heterogeneous liver iron distribution has been reported more frequently. These data suggest that the more pronounced cause of the T2* variability could not be attributed to an uneven iron distribution but to susceptibility artifacts.
Our finding is in contrast with the previously described histological data (14, 16, 24). However, our study population is different, considering a homogenous and large cohort of TM patients. Moreover, the resolution of our technique may not be adequate to detect the iron heterogeneity showed by histology.
The segmental and the mean hepatic T2* measurements were strongly correlated with the single ROI values (r always greater than 0.97, Table 2) more extensively confirming the datum showed in a preliminary paper where the liver T2* analysis was limited to a single slice (19).
The percentage deviation from single ROI value was negligible in segments VII and VIII. Single ROI measurement underestimates T2* values in the other segments (Table 2). The variations for these segments ranged from 12.6% to 17.5%. In fact, segments VII and VIII are the preferred locations in the single ROI analysis, due to the presence of a large area without blood vessels (10). The divergence between single ROI T2* values and mean hepatic T2* values was lower in patients with an abnormal single ROI T2* value.
In conclusion, in the thalassemia major patients T2* variations in liver are low and likely due to the artifact effects and measurement variability. The commonly used single ROI approach may slightly underestimate liver iron quantification due to susceptibility artifacts when the ROI was placed over segments VII and VIII, to date preferred locations in clinical practice. Thus, the single ROI approach can be safely used in the clinical arena, taking care to avoid the susceptibility artifact effects, found most frequently in segments VII and VIII.
We thank the following colleagues from the Italian thalassemia centers involved in the MIOT network: M. C. Putti (Università /Azienda Ospedaliera, Padova), A. Zuccarelli (Ospedale Civile, Olbia), C. Tassi (Policlinico S. Orsola, Bologna), S. Campisi (A. O. Umberto I, Siracusa), A. Filosa (Ospedale Cardarelli, Napoli), A. Spasiano (Ospedale Cardarelli, Napoli), D. D'Ascola (A.O. Bianchi-Melacrino-Morelli, Reggio Calabria), T. Casini (Ospedale Meyer, Firenze), M. A. Romeo (Azienda Policlinico, Catania), B. Piraino (Policlinico G. Martino, Messina), C. Gerardi (Ospedali Civili Riuniti, Agrigento), A. Quarta (Ospedale A. Perrino, Brindisi), M. E. Lai (Ospedale Microcitemico, Cagliari), M. G. Bisconte (Presidio Ospedaliero Annunziata, Cosenza), G. Giuffrida (Ospedale Ferrarotto, Catania), A. Quota (Ospedale V. Emanuele III, Gela), M. Rizzo (Ospedale Sant'Elia, Caltanisetta), R. Giugno (A. O. Gravina, Caltagirone), M. Furbetta (Policlinico Monteluce, Perugia), C. Fidone (A. O. Civile, Ragusa), M. P. Smacchia (Policlinico Umberto 1, Roma), G. Secchi (Azienda USL # 1, Sassari), S. Armari (Azienda Ospedaliera, Legnago), and M. L. Boffa (Ospedale S. Giovanni Bosco, Napoli). Thanks to the industrial sponsorships (Chiesi, and GE Healthcare) that gave their “no-profit support” for the MIOT study. Finally, we thank all patients for their cooperation, Claudia Santarlasci for skillful secretarial work, and Alison Frank for her assistance in editing this manuscript.