Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

A Practical Approach for a Wide Range of Liver Iron Quantitation Using a Magnetic Resonance Imaging Technique

A Practical Approach for a Wide Range of Liver Iron Quantitation Using a Magnetic Resonance... Hindawi Publishing Corporation Radiology Research and Practice Volume 2012, Article ID 207391, 9 pages doi:10.1155/2012/207391 Clinical Study A Practical Approach for a Wide Range of Liver Iron Quantitation Using a Magnetic Resonance Imaging Technique 1 2 2 1 3 Ping Hou, Uday R. Popat, Richard J. Lindsay, Edward F. Jackson, and Haesun Choi Department of Imaging Physics, The University of Texas MD Anderson Cancer Center, 1515 Holcombe Boulevard, Houston, TX 77030, USA Department of Stem Cell Transplantation and Cellular Therapy, The University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA Department of Diagnostic Radiology, The University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA Correspondence should be addressed to Ping Hou, ping.hou@mdanderson.org Received 27 August 2012; Revised 12 November 2012; Accepted 13 November 2012 Academic Editor: Paul Sijens Copyright © 2012 Ping Hou et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The goal of this study is to demonstrate a practical magnetic resonance imaging technique for quantifying a wide range of hepatic iron concentration (HIC) for hematologic oncology patients with transfusion iron overload in a routine clinical setting. To cover awiderange of T values from hematologic patients, we used a dual-acquisition method with two clinically available acquisition protocols on a 1.5T MRI scanner with different ΔTEs to acquire data in two breath-holds. An in-house image postprocessing software tool was developed to generate T , iron maps, and water and fat images, when fat is presented in the liver. The resulting iron maps in DICOM format are transferred to the institutional electronic medical record system for review by radiologists. The ∗ ∗ measured liver T values for 28 patients ranged from 0.56± 0.13 to 25.0± 2.1 milliseconds. These T values corresponded to HIC 2 2 values ranging from 1.2 ± 0.1mg/g to 45.0 ± 10.0 mg/g (dry weight). A moderate correlation between overall serum ferritin levels and R was found with a correlation coefficient of 0.83. Repeated phantom scans confirmed that the precision of this method is better than 4% for T measurements. The dual- acquisition method also improved the ability to quantify HIC of the patients with hepatic steatosis. 1. Introduction used to confirm the diagnosis of the iron overload disease and to monitor it during the course of treatment. However, Patients with transfusion-dependent hematologic malignan- liver biopsy is an invasive procedure with significant risk in cies, such as myelogenous leukemia, can often be treated patients with low platelet counts, and the technique to quan- successfully with stem cell transplantation [1–3]. However, titate iron in liver tissue is not widely available. Because of the after stem cell transplantation, transfusional hemosiderosis limitations of liver biopsy, a noninvasive method of hepatic may develop. If left untreated, the cumulative effects of iron iron detection and quantification, such as MRI, has been overload may lead to significant morbidity or even mortality. investigated by many researchers [4–24]. It has been shown The liver is the first and foremost organ affected by iron that the transverse relaxation times T and T are inversely overload, and hepatic iron overload has been associated with proportional to HIC [5], which means that the transverse the development of hepatitis, hepatic fibrosis, and cirrhosis. relaxation rates R and R , which are the reciprocals of T 2 2 The heart can also be affected by iron overload, resulting and T , respectively, are directly proportional to HIC. Using in heart failure. Thus, it is important to accurately diagnose various MRI techniques, several investigators have measured and adequately treat iron overload, especially in patients with HIC from T decay curve by spin echo (SE) method [7, 8, 14] transfusion-dependent hematologic malignancies. 2 and T decay curves generated by gradient recalled echo Since much of the body’s excess iron is deposited in the liver, hepatic iron concentration (HIC) is often used as a sur- (GRE) sequences [8–10, 20–25]. Another method widely rogate for the total body iron load. Liver biopsy is currently used in Europe [11–18] is based on signal ratio of liver over 2 Radiology Research and Practice reference tissue, using multiple images such as T -,T -, 2.2. MRI Techniques. All MRI scans were performed using proton-density- (PD-) weighted images from GRE and SE an 8-channel Torso cardiac coil on a 1.5-T MRI scanner sequences. In general, T method takes longer time to scan, (General Electric Medical Systems, Milwaukee, WI, USA), each breath hold can only generate one image at one echo with a maximum gradient of 40 mT/m and a slow rate time. T method is faster, more sensitive to iron con- of 150 mT/m/sec. After acquiring standard abdominal MR centration and generates multiple images at different echo images, such as T W in-out of phase and respiratory time within one breath hold. In addition, Wood et al. [8] triggered T W images, we acquired MEGE T -weighted demonstrated that the T and T method generated com- images through the largest section of the liver in the axial parable HIC value. The HIC values measured from MRI plane away from major vessels. An oblique coronal was also R maps have further been validated by other investigators acquired in order to check liver heterogeneity. The lung was against the HIC values obtained from the chemical analysis excluded on the axial plane as much as possible to avoid of needle biopsy specimens, which is considered the standard introducing potential susceptibility artifacts. method for measuring HIC [8, 9]. Wood’s method used From curve fitting point of view, we should acquire many only one protocol with relatively short TE (maximum TE = echoes with short ΔTE (less than 1.0 msec) to cover the 4.8 msec) and short ΔTE (0.8 msec), which may result in entire T decay curve so that T value could be accurately greater uncertainty in long T data acquisition. determined. The clinically available MEGE sequence can only acquire the maximum of sixteen echoes on our scanner, In oncologic patient, accumulation of fat in the liver fol- which limits the range of T that can be measured accurately lowing extensive chemotherapy is often a clinical concern. 2 if inadequate ΔTE (either too high or too low) is used. Using magnitude images acquired by multiecho gradient For a given patient, without prior knowledge of T or iron echo sequence (MEGE), several researchers have measured concentration, it is difficult to select the “right” ΔTE without proton density fat fraction (PDFF) [26–28]. And they “trial and error.” To overcome this difficulty, we used two pointed out that T would complicate fat quantification. In MRI acquisition protocols with different ΔTEs for each all of these published studies, four to eight echoes are patient to acquire data in two breath-holds. Of the two acquired, with each echo set at the time of water-fat in and measurements, the one that had better correlation and out of phases so that fat fraction can be measured accurately. smaller chi-square value was selected for subsequent T and However, short T is not measured adequately because the 2 iron map generation. minimum TE and ΔTE is not short enough to catch the fast The imaging parameters were optimized to include the decay MR signal from high iron overloaded liver. following considerations: (1) the shortest TE and a reason- In clinical practice, we encounter patients with a wide able ΔTE for curve fitting; (2) a high enough signal-to-noise range of HIC, from severe iron overload with very short ∗ ratio (SNR) to allow the liver to be differentiated from back- T (<2 msec) images to moderate-low iron overload with ∗ ground noise for very fast T decay cases; (3) a short enough relative long T . Currently, there is no “one size fits all” scan time to allow a single breath-hold for most patients. The technique for liver iron quantification. The aim of this long TE protocol was for long T (>3 msec) cases using a work is to (1) optimize clinically available MR acquisition ∗ single shot, TR = 39.1 msec, 12 echoes, a unipolar readout protocols to cover a wide range of T decay; (2) develop gradient, TE1 = 1.448 msec, and ΔTE = 2.336 msec. The short postprocessing tools to generate reliable HIC maps; (3) TE protocol was for short T (<3 msec) cases using double generate and display the HIC map in a timely manner for shots, 8 echoes/shot, TR = 20 msec, a bipolar readout gradi- radiologists to review. Our method, employing two MEGE ent, TE1 = 1.448 msec, and ΔTE = 0.636 msec. Each protocol acquisition protocols with different TE coverage to cover a ∗ had a scan time of less than 16 seconds, with the longest TEs wide range of T , is an extension of Wood’s work. It not ∗ being 27 msec and 11 msec, respectively. The other imaging only measures wide range of clinically relevant T , but also parameters, including matrix size (256 × 192), field of view allows us to simultaneously measure HIC and fat fraction for (38 cm), flip angle (25 ), receiver bandwidth (125 kHz), slice patients with hepatic steatosis. thickness (10 mm), and averages (2), were kept the same in all protocols. Among the two acquisitions, only one of the T fitted images with the higher correlation coefficient R of the 2. Materials and Methods fit, was selected to generate HIC map for clinical use. 2.1. Patient Population. This retrospective study was per- To validate the reproducibility of our method and test formed with the approval of our institutional review board, the stability of the system, a FerriScan phantom was used to measure T values with the same protocol for patients which waived the requirement for informed consent. We identified 28 consecutive patients with high ferritin who in every three to four months over a period of one year. underwent hematopoietic stem cell transplant at our insti- This phantom consists of 15 vials of aqueous manganese tution. Of these 28 patients, 19 had acute myelogenous chloride (MnCl ) solutions (0 mM to 3.2 mM) in 10 mM leukemia (AML), 3 had myelodysplastic disease (MD), 3 hydrochloric acid, which provide the T values from 3.0 msec to 40.0 msec, similar to the range of liver T values had chronic myelogenous leukemia (CML), and 3 had myelofibrosis. Clinically, it is believed that the amount of for patients. It is the same type of phantom used by Pierre et ferritin in blood reflects the amount of iron stored in the liver al. [7], who validated their technique for HIC measurement using the phantom and biopsy data. Their single echo SE [13]. Therefore, the serum ferritin level was measured within one month of the MRI scan in all patients. based technique would take 20–30 minutes to scan. Radiology Research and Practice 3 ∗ ∗ 2.3. Postprocessing MR Images . The T , R , and hepatic iron 3. Results 2 2 maps were processed off-line using an in-house developed For all 28 patients, the average T values ranged from 0.56± software tool based on MATLAB (Mathworks, Inc., Natick, 0.13 msec to 25.0 ± 2.1 msec. The corresponding HIC values MA, USA). The T curve was fitted to a monoexponential calculated from (1)–(3)werefrom1.2 ± 0.10 mg/g to 45.0 ± decay with three parameters using the Levenberg-Marquardt 10.0 mg/g dry weight. nonlinear least-squares method: Figure 1 shows a representative fitted signal versus time curves with long T value. Dual data acquisition method S(t) = A exp − + C,(1) was applied to the same patient, with fitted curves as shown in Figures 1(a) and 1(b),respectively. The T decay curve where A is the signal amplitude at time 0, with its initial fit acquired from the long TE protocol resulted in the better value set to 1.5 or 5.0 times of the maximum signal of the fit for the long T case, as demonstrated in Figure 1(a). ∗ ∗ ∗ first echo for each pixel for long T or short T cases; T is Therefore, the image from the long TE protocol was selected 2 2 2 initially set to 10 msec; and C represents noise whose initial to generate an iron map. Figure 2 is an example of a very fit value was set to the background noise. The background short T case with dual acquisition method in (a) and (b), noise was measured from a small ROI in the frequency respectively. The data acquired by the short TE protocol (for direction; its mean was used as a threshold to roughly mask short T values) provided better fit (b), as demonstrated by the image out before fitting. When there is fat in the liver, the its correlation coefficient and chi-square value. Therefore, in MRI signal in the liver can be represented by this case, images acquired by the short TE protocol were used to generate an iron map. In general, the more points sampled near the part of the curve with larger curvature, the more S(t) = W + F exp(iωt) exp − + C,(2) T reliable the generated results. For this reason, the longer TE and ΔTE spans were used for the long T fit, and the shorter where W is the water signal amplitude, F is the fat signal TE and ΔTE spans were used for the short T fit. As shown amplitude, and ω is the water-fat frequency difference that ∗ in Figure 2, the short T had faster decay and generated very is about 220 Hz at 1.5 T. Other parameters are the same as in low-intensity liver images. The signal intensity after the first (1). echo acquired by the long TE protocol, or after the third After fitting T curve using (1)or(2), the iron value, Fe echo acquired by the short TE protocol, was very close to the in mg/g (dry weight), was calculated based on a previously background noise and thus was much weaker than the signal validated linear relationship between iron and R value in from the long T (Figure 1)case. Hz [8]: Figure 3 displays a typical HIC map and its histogram of a 31-year-old man with acute myelogenous leukemia who F = 0.0254 · R +0.202. e (3) had very high levels of iron deposition. Figure 3(a) is the HIC map of the segmented liver, and the corresponding his- Both region of interest (ROI) and pixelwise fitting were togram with normal distribution fit is shown in Figure 3(b). implemented for a quick evaluation and T map generation. A combination of the HIC map and liver-only histogram A small ROI was drawn to plot the signal intensity versus provide local and global measurements of the response of the time curve for a quick review. If there was no signal liver to the treatment. oscillation, that means there was no fat in the liver, (1) Figure 4 is an example of T decay curves from a was used for curve fit; otherwise (2) was used for curve fit. hepatic steatosis patient acquired by our dual acquisition To reduce noise, the original images were smoothed by a ∗ method. There was no water-fat signal oscillation from 3 × 3 window kernel prior to curve fitting for short T the long TE acquisition protocol in Figure 4(a).However, cases. Since image intensity is not uniformly distributed due water-fat peaks were clearly demonstrated with the data to liver heterogeneity, tissue susceptibility, and surface coil acquired by the short TE protocol, as shown in Figure 4(b). sensitivity, local ROI measurements cannot represent the Figure 5 demonstrates that the long TE acquisition protocol global iron overload of the liver. After the pixelwise T curve is necessary to measure mild iron concentration as well as fitting, we segmented out the liver from the entire image PDFF for a hepatic steatosis patient. to form liver-only HIC map and generated a histogram of Shown in Figure 6 are the T measurement results (b) the liver-only HIC map with a normal distribution fit. This from a FerriScan Phantom consisting of fifteen different histogram shape, mean and standard deviations were used liquid tubes (a) with different T values. Three data sets were for longitudinal HIC comparison for patient followups. acquired with the same protocol for patient scan in a year The iron map generated by (3) was saved as a Digital with 3-4 months apart. Imaging and Communications in Medicine (DICOM) Figure 7 presents the relationship between R and ferritin image, its DICOM header was derived from original MRI for all patients. A paired Student’s t test was run between R image with different description and series number, and and ferritin values, with a correlation of 0.83 and P = 0.0001. was transferred to the institutional electronic medical record system, ClinicStation (Datalign, Inc., Houston, TX, USA), 4. Discussion for measurement and interpretation by radiologists. Radiol- ogists can obtain HIC values directly from the iron map by We have developed a simple and practical MRI strategy drawing an ROI. that provides the absolute HIC. By combining two different 4 Radiology Research and Practice Long TE protocol Long TE protocol ∗ 24 T = 23.16 ± 4.43 ms 140 2 R = 0.9963, χ = 8.561 ∗ T = 0.71 ± 0.66 ms R = 0.974, χ = 0.5356 60 14 0 510 15 20 25 30 0 5 10 15 20 25 30 Time (ms) Time (ms) Original Original 3-parameter fit 3-parameter fit (a) (a) Short TE protocol Short TE protocol T = 16.91 ± 22.9 ms R = 0.9514, χ = 11.0521 T = 0.57 ± 0.03 ms R = 0.9966, χ = 0.0855 90 18 02468 10 12 02468 10 12 Time (ms) Time (ms) Original Original 3-parameter fit 3-parameter fit (b) (b) ∗ ∗ Figure 2: Results of T curve fit from a patient with short T 2 2 ∗ ∗ Figure 1: Results of T curve fit from a patient with long T values 2 2 values by dual acquisition method (Figure 1(a)). Data was acquired by dual acquisition method. (a). Data was acquired by the long TE by the long TE protocol (Figure 1(b)). The same patient data was protocol. (b) the same patient data was acquired by the short TE acquired by the short TE protocol, there were more data points protocol. It is obvious for this case that the T fit error is increasing in the curvature area. It is obvious that data acquired by short TE with shorter TE coverage. The long TE data acquisition is necessary ∗ protocol generated better fit for short T case. for mild iron load patient. MEGE data acquisition protocols and an in-house image ΔTE and longer TE coverage data acquisition method should postprocessing tool, we cannot only measure a wide range of be used. Inadequate data acquisition protocol would result clinically relevant iron loads, from 1.2 mg/g to 45 mg/g (dry in great uncertainties in the value of HIC, as demonstrated weight), but also quantify water and fat simultaneously with in Figures 1 and 2. In this study, because the MRI scans confidence. This technique is now routinely used to evaluate were performed without prior knowledge of each patient’s HIC in all patients with suspected iron deposition disease in HIC, dual acquisition method covering different ranges of our institution. TE span was used. Since each acquisition took less than For patients with high levels of hepatic iron deposition 16 seconds, we were able to scan each patient with both (e.g., HIC > 15 mg/g), the T value was very short, and thus protocols (for different ΔTEs covering the entire range of a short ΔTE data acquisition method should be used. For possible T values) in two breath-holds. The data acquired patients with low levels of hepatic iron deposition (e.g., with the longer TE span generated better fits for the long T HIC < 3 mg/g), the T value was long and thus a relative long curve (Figure 1(a)), and the data acquired with the shorter Intensity Intensity Intensity Intensity Radiology Research and Practice 5 Long TE protocol T = 3.37 ± 0.36 ms R = 0.99, χ = 2.5891 0 5 10 15 20 25 30 Time (ms) (a) Original 3-parameter fit HIC histogram (a) HIC = 30.2 ± 5.3 mg/g 50 Short TE protocol Water = 71.31, fat = 42.88 35 T = 2.7 ± 0.11 ms R = 0.9982, χ = 0.2811 0 1020304050 HIC (mg/g) 02468 10 12 (b) Time (ms) Figure 3: A 31-year-old man with acute myelogenous leukemia, Original sixteen months after bone marrow transplantation. (a) is the HIC 3-parameter fit map of the segmented liver. (b) is its corresponding histogram with (b) normal distribution fit. This patient had very high iron deposition with mean HIC value of 30.20 mg/g (T < 1.0msec). Figure 4: An example of T curve fit from a liver steatosis patient with short T values and by dual acquisition method. (a) Data was acquired by the long TE protocol. The water and fat peaks were totally missed because data was sampled too slowly. (b) The same patient data was acquired by the short TE protocol with faster TE span generated better fits for the short T curves sample (shorter ΔTE). Water and fat peaks were caught and data (Figure 2(b)). was fitted with water and fat components. Again data acquired by A popular method to quantify liver iron by MR in the short TE protocol generated better fit for short T case. Europe is based on signal ratio of liver over reference tissue [11–18]fromdifferent images. Recently, this method has been further developed into a web-based iron measurement (URennes) [14] where proton-density-, T -, T -, and T - was 61.4% with a tendency of overestimate overload. They 1 2 2 weighted images have to be acquired based on their protocol, also found that the iron concentration by this method was and certain ROIs have to be measured in the liver and less reliable from 60 μmol/g to 170 μmol/g, corresponding to reference tissues as the inputs to calculate an adequate 3.35 mg/g to 9.5 mg/g. Another limitation of this method was iron value. Gandon et al. [16] also validated this technique that it could only measure iron up to 350 μmol/g (19.5 mg/g). with biopsy data on 1.5T system, declaring that it could This method was developed early in late 1990s [12] when be used on various MR scanners. Most recently, Castiella MEGE sequence was not commercially available. et al. [18] reevaluated this method using the data from Wood et al. [8] studied more than 100 patients for multiple institutions with different scanners in the period of HICs by T (SE), T (GRE) method. Among them, 21 were 2 2 1999 to 2006. They investigated the accuracy of this method liver biopsied and validated. They reported that the GRE with liver biopsy and found that the diagnostic accuracy sequence was able to adequately measure T values in the Pixels Intensity Intensity 6 Radiology Research and Practice Long TE protocol Water = 126.54, fat = 8.9 90 ∗ T = 20.64 ± 3.05 ms R = 0.9981, χ = 2.66 0 5 10 15 20 25 30 Time (ms) Original (a) 3-parameter fit Figure 5: An example of T decay curve fit from a liver steatosis patient with less iron deposition. The long TE protocol was adequate to catch the water and fat peak while minimum fit error for long T decay was achieved. entire range of iron overload. This method has recently been further validated by other investigators [9]. In our study, although the minimum TE was 1.448 msec, we were stillabletomeasure T values less than 1.0 msec from 8- echo, double-shot interleaved sequences, as demonstrated in Figure 2(b), where the SNR dropped to the background level at the fourth echo. This T value corresponds to a HIC value over 40 mg/g for that specific ROI, which is extremely S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13 S14 S15 high for a patient with acute myeloid leukemia [7, 8]. For Sample such high iron overload, raw images need to be smoothed to ∗ Scan 1 remove some noises. It is known that the T value for normal Scan 2 liver is ∼24 msec [20–22]. Therefore, we have demonstrated Scan 3 that our technique is able to measure the entire range of (b) possible clinical HIC values using the clinically available MRI sequence. With the maximum receiver bandwidth used Figure 6: Results of phantom validation. (a) is the image of in our protocol, the first TE could be reduced further by the FerriScan phantom consisting of fifteen different tubes with using a 128 × 128 or even 64 × 64 acquisition matrix. This different solutions (different T values). The sample numbers are might be beneficial for SNR for cases with very short T defined on the left. (b) is the T measurement from three scans in a 2 2 values (<1.0 msec). Meanwhile, Figure 1 demonstrates the year using the same clinical protocols. Consistent measurement was importance of long TE acquisition for long T decay case. demonstrated in this plot. Short TE acquisition for long T decay case would result in larger measurement uncertainty, a limitation of Wood’s method. in Figure 2(b), we did observe significant noise from data The slope and intercept for the linear equation (3) acquisition, especially close to the high tissue susceptibility between HIC and R under 1.5-T was determined by Wood area and low coil sensitivity area. For patients with high et al. [8] using biopsy results. It was also calibrated and HIC values, having an HIC map is necessary to view the confirmed with biopsy values by Hankins et al. [9]. In our image nonuniformity, an example is shown in Figure 3(a). study, we used Wood et al.’s results and did not validate However, this HIC image map is not enough to track our HIC values against those from biopsies. In addition, the overall mean HIC value and standard deviation. A Wood pointed out that their calibration was validated from liver-only HIC histogram with normal distribution fit, as 1.3 mg/g to 32.9 mg/g dry weight. There was an outlier shown in Figure 3(b), can provide the mean and standard with HIC value of 57.8 mg/g, it was removed from their deviations of the liver’s iron concentration. We have found statistic calibration. With the minimum TE of 1.448 msec that a combination of HIC image map and its histogram in our dual acquisition method, although we can measure distribution is extremely useful for patient followup after and fit the T value as short as 0.57 msec, as demonstrated treatment. Intensity T (ms) 2 Radiology Research and Practice 7 R versus ferritin observed curve fit improvement. Therefore, in our study, we assumed that the T decay was identical for water and fat signal (2). The independent T decay model needs to be further studied with a larger number of patients. Researchers [21, 22, 30, 31] have demonstrated that iron measurement from T method has high reproducibility and inter-MRI scanner agreement. Using FerriScan phantom, we have confirmed the reproducibility (Figure 6)of T measurements for all samples. The average error was 2.1% with the maximum error of 3.9%. The largest error mainly came from nonuniform image due to phantom/slice position between each scans, seen in sample 3 and 6. The consistent measurement from fixed phantom samples further demon- strated that our technique is quite promising for patient followups. In our study, we have found that consistency in data acquisition and analysis is vital for patient followups. As mentioned earlier, there was a reasonably good corre- lation between the R and ferritin values (Figure 7)inour study. Although the serum ferritin values have been used as a 0 5000 10000 15000 marker for body iron amount [19], some researchers [9, 10] Ferritin (ng/ml) have shown that the ferritin values do not necessarily reflect y = 127.97 + 0.10375x, R = 0.82969 the actual total body iron burden. Therefore, using serum R (Hz) ferritin to represent iron load is still a debatable topic. On ∗ the other hand, HIC as measured by an MEGE sequence Figure 7: Scattergram of R values calculated from MEGE MRI does represent the actual hepatic iron burden, as supported acquisition versus serum ferritin values. A good correlation between by several biopsy reports [7–9]. Using the dual acquisition R and ferritin is clearly demonstrated. method described in this study, the MEGE sequence can be used in the clinical decision-making with higher confidence. Our dual acquisition method cannot only cover wide Image quality will affect the iron measurements. First, range of clinical relevant T decay (all iron deposition), but susceptibility artifacts will affect our measurement of HIC. also quantify iron accurately for the patient with superim- Shimming could help to reduce some B0 inhomogeneity, and posed steatosis, as demonstrated in Figure 4. For hepatic positioning the patient carefully so that liver is not close to steatosis patient with less iron deposition, the long TE the edge of coil will help to generate more uniform image. protocol can detect water and fat oscillation, as shown in Since the whole HIC map is generated, radiologist can always Figure 5. Meanwhile, the short TE protocol with shorter ΔTE put an ROI to the most homogeneous part to measure HIC value. The histogram from liver-only HIC provides overall can detect the water and fat oscillation for patient with high iron deposition.Itiswellknown thatwater andfat will be mean and standard deviations, a useful distribution to mon- in phase at time 4.6 msec and out of phase at 2.3 msec on itor treatment. Meloni et al. [25] have studied single-slice versus multislice HIC distribution of HIC by T method and 1.5 T system. By setting the echo time at the multiple time of 2.3 msec, Sirlin and Reeder [26–28] studied fat concluded that the single-slice measurement is enough in the quantification by measuring PDFF using an MEGE sequence. clinical application as long as major vessels are avoided. Sec- Their methods can quantify fat and water, as well as generate ond, patient’s motion affects the T measurement, breathing R map with four to eight echoes. Their main purpose was resulting in severe nonuniform artifact in the liver image. fat assessment; therefore, the minimum TE and ΔTE were Breath instruction should be delivered clearly before data not short enough to quantify high iron overload cases. For acquisition and patients need to be trained to hold breath. extremely high iron overload such as T < 1.0 msec, steatosis Our technique has some limitations. One limitation is the lack of validation of patients with combined iron and could not be detected by image technique because water and fat peak may not be MRI visible. In our study, no fatty liver fat deposition in the liver, especially in high HIC (T < was detected with T value less than 2.0 msec. To improve 2.0 msec) and high PDFF cases. The second limitation is that the TEs are not set at the multiple of 2.3 msec for long the accuracy of HIC and fat measurement, a better pulse sequence is needed with the minimum TE and ΔTE less than TE data acquisition protocol. This could result in PDFF 1.0 msec, the maximum TE greater than 20.0 msec, and some underestimation for minor iron load patient, such as the case of the echo should be set at multiple times of 2.3 msec. in Figure 5. Another minor problem is that the multiecho Chebrolu et al. [29] pointed out that the independent T -weighted images have to be transferred from the MR scanner to a local workstation for postprocessing, and it T decay for water and fat model may improve accuracy of water-fat quantification using a 3D MEGE sequence, takes several minutes for an HIC map to be generated. Since particularly for high fat fraction and short T case. We have the postprocessing on a local workstation is automatic, it is still practical. It would be ideal to generate the HIC map used independent T model to fit the signal decay curve. For our limited number of steatosis patients, we have not immediately after image acquisition on the MRI scanner. R (Hz) 2 8 Radiology Research and Practice 5. Conclusion Roentgenology, vol. 168, no. 5, pp. 1205–1208, 1997. [13] M. Queiroz-Andrade, R. Blasbalg, C. D. Ortega et al., “MR We have developed a simple, quick, and reliable method that imaging findings of iron overload,” Radiographics, vol. 29, no. can noninvasively measure hepatic iron overload using a 6, pp. 1575–1589, 2009. [14] Y. Gandon, “Iron, liver and MRI,” University of Rennes, http:// standard 1.5-T MRI scanner. By combining different MEGE www.radio.univ-rennes1.fr/Sources/EN/Hemo.html. data acquisition strategies and an in-house image postpro- [15] J. M. Alustiza, ´ J. Artetxe, A. Castiella et al., “MR quantification cessing tool, our technique cannot only measure a wide of hepatic iron concentration,” Radiology, vol. 230, no. 2, pp. range of clinically relevant HIC values from 1.2 mg/g to 479–484, 2004. greater than 45.0 mg/g dry weight, but also quantify fat/water [16] Y. Gandon, D. Olivie, ´ D. Guyader et al., “Non-invasive assess- fraction simultaneously. It generates an absolute HIC map ment of hepatic iron stores by MRI,” The Lancet, vol. 363, no. and transfers it to the electronic medical record system, 9406, pp. 357–362, 2004. which makes it convenient for the physicians to diagnose and [17] A. W. Olthof,P.E.Sijens, H. G. Kreeftenberg,P.Kappert,E. measure hepatic overload. J. van der Jagt, and M. Oudkerk, “Non-invasive liver iron concentration measurement by MRI: comparison of two validated protocols,” European Journal of Radiology, vol. 71, References no. 1, pp. 116–121, 2009. [18] A. Castiella, J. M. Alustiza, ´ J. I. Emparanza, E. M. Zapata, B. [1] T. Robak and A. Wierzbowska, “Current and emerging ther- apies for acute myeloid leukemia,” Clinical Therapeutics, vol. Costero, and M. I. D´ıez, “Liver iron concentration quantifi- cation by MRI: are recommended protocols accurate enough 31, no. 2, pp. 2349–2370, 2009. [2] J. J. Cornelissen, W. L. J. Van Putten, L. F. Verdonck et al., for clinical practice?” European Radiology,vol. 21, no.1,pp. 137–141, 2011. “Results of a HOVON/SAKK donor versus no-donor analysis of myeloablative HLA-identical sibling stem cell transplanta- [19] A. W. Olthof, P. E. Sijens, H. G. Kreeftenberg et al., “Correla- tion between serum ferritin levels and liver iron concentration tion in first remission acute myeloid leukemia in young and middle-aged adults: benefits for whom?” Blood, vol. 109, no. determined by MR imaging: impact of hematologic disease 9, pp. 3658–3666, 2007. and inflammation,” Magnetic Resonance Imaging, vol. 25, no. 2, pp. 228–231, 2007. [3] D. Vicente, T. Lamparelli, F. Gualandi et al., “Improved out- come in young adults with de novo acute myeloid leukemia in [20] H. Chandarana, R. P. Lim, J. H. Jensen et al., “Hepatic iron deposition in patients with liver disease: preliminary expe- first remission, undergoing an allogeneic bone marrow trans- plant,” Bone Marrow Transplantation, vol. 40, no. 4, pp. 349– rience with breath-hold multiecho T2 -weighted sequence,” American Journal of Roentgenology, vol. 193, no. 5, pp. 1261– 354, 2007. [4] D.D.Stark,N.M.Bass, andA.A.Moss, “Nuclear magnetic 1267, 2009. [21] M. Westwood, L. J. Anderson, D. N. Firmin et al., “A single resonance imaging of experimentally induced liver disease,” Radiology, vol. 148, no. 3, pp. 743–751, 1983. breath-hold multiecho T2 cardiovascular magnetic reso- nance technique for diagnosis of myocardial iron overload,” [5] R. Engelhardt, “Liver iron quantification: studies in aqueous iron solutions, iron overloaded rats, and patients with heredi- Journal of Magnetic Resonance Imaging, vol. 18, no. 1, pp. 33– 39, 2003. tary hemochromatosis,” Magnetic Resonance Imaging, vol. 12, no. 7, pp. 999–1007, 1994. [22] L. J. Anderson, S. Holden, B. Davis et al., “Cardiovascular ∗ ∗ [6] Z.J.Wang, J. C. Haselgrove,M.B.Martinetal., “Evaluationof T2 (T2 ) magnetic resonance for the early diagnosis of myocardial iron overload,” European Heart Journal, vol. 22, no. iron overload by single voxel MRS measurement of liver T2,” Journal of Magnetic Resonance Imaging, vol. 15, no. 4, pp. 395– 23, pp. 2171–2179, 2001. [23] S. D. Chirnomas, M. Geukes-Foppen, K. Barry et al., “Practical 400, 2002. [7] T. G. St Pierre, P. R. Clark, W. Chua-Anusorn et al., “Nonin- implications of liver and heart iron load assessment by T2 -MRI in children and adults with transfusion-dependent vasive measurement and imaging of liver iron concentrations using proton magnetic resonance,” Blood, vol. 105, no. 2, pp. anemias,” American Journal of Hematology, vol. 83, no. 10, pp. 781–783, 2008. 855–861, 2005. [8] J. C. Wood, C. Enriquez, N. Ghugre et al., “MRI R2 and [24] V. Positano, B. Salani, A. Pepe et al., “Improved T2 assess- ment in liver iron overload by magnetic resonance imaging,” R2 mapping accurately estimates hepatic iron concentration in transfusion-dependent thalassemia and sickle cell disease Magnetic Resonance Imaging, vol. 27, no. 2, pp. 188–197, 2009. [25] A. Meloni, A. Luciani, V. Positano et al., “Single region of patients,” Blood, vol. 106, no. 4, pp. 1460–1465, 2005. ∗ ∗ [9] J.S.Hankins,M.B.McCarville, R. B. Loeffler et al., “R2 interest versus multislice T2 MRI approach for the quantifi- magnetic resonance imaging of the liver in patients with iron cation of hepatic iron overload,” Journal of Magnetic Resonance Imaging, vol. 33, no. 2, pp. 348–355, 2011. overload,” Blood, vol. 113, no. 20, pp. 4853–4855, 2009. [10] A. W. K. Leung,W.C.W.Chu,W.W.M.Lam,V.Lee,and C. [26] H. Yu, C. A. McKenzie, A. Shimakawa et al., “Multiecho recon- struction for simultaneous water-fat decomposition and T2 K. Li, “Magnetic resonance imaging assessment of cardiac and liver iron load in transfusion dependent patients,” Pediatric estimation,” Journal of Magnetic Resonance Imaging, vol. 26, no. 4, pp. 1153–1161, 2007. Blood and Cancer, vol. 53, no. 6, pp. 1054–1059, 2009. [11] Y. Gandon, D. Guyader, J. F. Heautot et al., “Hemochromato- [27] M. Bydder,T.Yokoo,G.Hamiltonetal., “Relaxationeffects in the quantification of fat using gradient echo imaging,” sis: diagnosis and quantification of liver iron with gradient- echo MR imaging,” Radiology, vol. 193, no. 2, pp. 533–538, Magnetic Resonance Imaging, vol. 26, no. 3, pp. 347–359, 2008. [28] S. B. Reeder, I. Cruite, G. Hamilton, and C. B. Sirlin, “Quanti- [12] O. Ernst, G. Sergent, P. Bonvarlet, V. Canva-Delcambre, J. C. tative assessment of liver fat with magnetic resonance imaging ´ and spectroscopy,” Journal of Magnetic Resonance Imaging, vol. Paris, and C. L’Hermine, “Hepatic iron overload: diagnosis and quantification with MR imaging,” American Journal of 34, no. 4, 2011, spcone. Radiology Research and Practice 9 [29] V. V. Chebrolu, C. D. G. Hines, H. Yu et al., “Independent estimation of T 2 for water and fat for improved accuracy of fat quantification,” Magnetic Resonance in Medicine, vol. 63, no. 4, pp. 849–857, 2010. [30] M. A. Westwood, D. N. Firmin, M. Gildo et al., “Intercentre reproducibility of magnetic resonance T2 measurements of myocardial iron in thalassaemia,” International Journal of Cardiovascular Imaging, vol. 21, no. 5, pp. 531–538, 2005. [31] M. A. Westwood, L. J. Anderson, D. N. Firmin et al., “Inter- scanner reproducibility of cardiovascular magnetic resonance T2 measurements of tissue iron in Thalassemia,” Journal of Magnetic Resonance Imaging, vol. 18, no. 5, pp. 616–620, 2003. MEDIATORS of INFLAMMATION The Scientific Gastroenterology Journal of World Journal Research and Practice Diabetes Research Disease Markers Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 International Journal of Journal of Immunology Research Endocrinology Hindawi Publishing Corporation Hindawi Publishing Corporation http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 Submit your manuscripts at http://www.hindawi.com BioMed PPAR Research Research International Hindawi Publishing Corporation Hindawi Publishing Corporation http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 Journal of Obesity Evidence-Based Journal of Journal of Stem Cells Complementary and Ophthalmology International Alternative Medicine Oncology Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 Parkinson’s Disease Computational and Behavioural Mathematical Methods AIDS Oxidative Medicine and in Medicine Research and Treatment Cellular Longevity Neurology Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Radiology Research and Practice Hindawi Publishing Corporation

A Practical Approach for a Wide Range of Liver Iron Quantitation Using a Magnetic Resonance Imaging Technique

Loading next page...
 
/lp/hindawi-publishing-corporation/a-practical-approach-for-a-wide-range-of-liver-iron-quantitation-using-8QsThrKI6a
Publisher
Hindawi Publishing Corporation
Copyright
Copyright © 2012 Ping Hou et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
ISSN
2090-1941
eISSN
2090-195X
DOI
10.1155/2012/207391
Publisher site
See Article on Publisher Site

Abstract

Hindawi Publishing Corporation Radiology Research and Practice Volume 2012, Article ID 207391, 9 pages doi:10.1155/2012/207391 Clinical Study A Practical Approach for a Wide Range of Liver Iron Quantitation Using a Magnetic Resonance Imaging Technique 1 2 2 1 3 Ping Hou, Uday R. Popat, Richard J. Lindsay, Edward F. Jackson, and Haesun Choi Department of Imaging Physics, The University of Texas MD Anderson Cancer Center, 1515 Holcombe Boulevard, Houston, TX 77030, USA Department of Stem Cell Transplantation and Cellular Therapy, The University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA Department of Diagnostic Radiology, The University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA Correspondence should be addressed to Ping Hou, ping.hou@mdanderson.org Received 27 August 2012; Revised 12 November 2012; Accepted 13 November 2012 Academic Editor: Paul Sijens Copyright © 2012 Ping Hou et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The goal of this study is to demonstrate a practical magnetic resonance imaging technique for quantifying a wide range of hepatic iron concentration (HIC) for hematologic oncology patients with transfusion iron overload in a routine clinical setting. To cover awiderange of T values from hematologic patients, we used a dual-acquisition method with two clinically available acquisition protocols on a 1.5T MRI scanner with different ΔTEs to acquire data in two breath-holds. An in-house image postprocessing software tool was developed to generate T , iron maps, and water and fat images, when fat is presented in the liver. The resulting iron maps in DICOM format are transferred to the institutional electronic medical record system for review by radiologists. The ∗ ∗ measured liver T values for 28 patients ranged from 0.56± 0.13 to 25.0± 2.1 milliseconds. These T values corresponded to HIC 2 2 values ranging from 1.2 ± 0.1mg/g to 45.0 ± 10.0 mg/g (dry weight). A moderate correlation between overall serum ferritin levels and R was found with a correlation coefficient of 0.83. Repeated phantom scans confirmed that the precision of this method is better than 4% for T measurements. The dual- acquisition method also improved the ability to quantify HIC of the patients with hepatic steatosis. 1. Introduction used to confirm the diagnosis of the iron overload disease and to monitor it during the course of treatment. However, Patients with transfusion-dependent hematologic malignan- liver biopsy is an invasive procedure with significant risk in cies, such as myelogenous leukemia, can often be treated patients with low platelet counts, and the technique to quan- successfully with stem cell transplantation [1–3]. However, titate iron in liver tissue is not widely available. Because of the after stem cell transplantation, transfusional hemosiderosis limitations of liver biopsy, a noninvasive method of hepatic may develop. If left untreated, the cumulative effects of iron iron detection and quantification, such as MRI, has been overload may lead to significant morbidity or even mortality. investigated by many researchers [4–24]. It has been shown The liver is the first and foremost organ affected by iron that the transverse relaxation times T and T are inversely overload, and hepatic iron overload has been associated with proportional to HIC [5], which means that the transverse the development of hepatitis, hepatic fibrosis, and cirrhosis. relaxation rates R and R , which are the reciprocals of T 2 2 The heart can also be affected by iron overload, resulting and T , respectively, are directly proportional to HIC. Using in heart failure. Thus, it is important to accurately diagnose various MRI techniques, several investigators have measured and adequately treat iron overload, especially in patients with HIC from T decay curve by spin echo (SE) method [7, 8, 14] transfusion-dependent hematologic malignancies. 2 and T decay curves generated by gradient recalled echo Since much of the body’s excess iron is deposited in the liver, hepatic iron concentration (HIC) is often used as a sur- (GRE) sequences [8–10, 20–25]. Another method widely rogate for the total body iron load. Liver biopsy is currently used in Europe [11–18] is based on signal ratio of liver over 2 Radiology Research and Practice reference tissue, using multiple images such as T -,T -, 2.2. MRI Techniques. All MRI scans were performed using proton-density- (PD-) weighted images from GRE and SE an 8-channel Torso cardiac coil on a 1.5-T MRI scanner sequences. In general, T method takes longer time to scan, (General Electric Medical Systems, Milwaukee, WI, USA), each breath hold can only generate one image at one echo with a maximum gradient of 40 mT/m and a slow rate time. T method is faster, more sensitive to iron con- of 150 mT/m/sec. After acquiring standard abdominal MR centration and generates multiple images at different echo images, such as T W in-out of phase and respiratory time within one breath hold. In addition, Wood et al. [8] triggered T W images, we acquired MEGE T -weighted demonstrated that the T and T method generated com- images through the largest section of the liver in the axial parable HIC value. The HIC values measured from MRI plane away from major vessels. An oblique coronal was also R maps have further been validated by other investigators acquired in order to check liver heterogeneity. The lung was against the HIC values obtained from the chemical analysis excluded on the axial plane as much as possible to avoid of needle biopsy specimens, which is considered the standard introducing potential susceptibility artifacts. method for measuring HIC [8, 9]. Wood’s method used From curve fitting point of view, we should acquire many only one protocol with relatively short TE (maximum TE = echoes with short ΔTE (less than 1.0 msec) to cover the 4.8 msec) and short ΔTE (0.8 msec), which may result in entire T decay curve so that T value could be accurately greater uncertainty in long T data acquisition. determined. The clinically available MEGE sequence can only acquire the maximum of sixteen echoes on our scanner, In oncologic patient, accumulation of fat in the liver fol- which limits the range of T that can be measured accurately lowing extensive chemotherapy is often a clinical concern. 2 if inadequate ΔTE (either too high or too low) is used. Using magnitude images acquired by multiecho gradient For a given patient, without prior knowledge of T or iron echo sequence (MEGE), several researchers have measured concentration, it is difficult to select the “right” ΔTE without proton density fat fraction (PDFF) [26–28]. And they “trial and error.” To overcome this difficulty, we used two pointed out that T would complicate fat quantification. In MRI acquisition protocols with different ΔTEs for each all of these published studies, four to eight echoes are patient to acquire data in two breath-holds. Of the two acquired, with each echo set at the time of water-fat in and measurements, the one that had better correlation and out of phases so that fat fraction can be measured accurately. smaller chi-square value was selected for subsequent T and However, short T is not measured adequately because the 2 iron map generation. minimum TE and ΔTE is not short enough to catch the fast The imaging parameters were optimized to include the decay MR signal from high iron overloaded liver. following considerations: (1) the shortest TE and a reason- In clinical practice, we encounter patients with a wide able ΔTE for curve fitting; (2) a high enough signal-to-noise range of HIC, from severe iron overload with very short ∗ ratio (SNR) to allow the liver to be differentiated from back- T (<2 msec) images to moderate-low iron overload with ∗ ground noise for very fast T decay cases; (3) a short enough relative long T . Currently, there is no “one size fits all” scan time to allow a single breath-hold for most patients. The technique for liver iron quantification. The aim of this long TE protocol was for long T (>3 msec) cases using a work is to (1) optimize clinically available MR acquisition ∗ single shot, TR = 39.1 msec, 12 echoes, a unipolar readout protocols to cover a wide range of T decay; (2) develop gradient, TE1 = 1.448 msec, and ΔTE = 2.336 msec. The short postprocessing tools to generate reliable HIC maps; (3) TE protocol was for short T (<3 msec) cases using double generate and display the HIC map in a timely manner for shots, 8 echoes/shot, TR = 20 msec, a bipolar readout gradi- radiologists to review. Our method, employing two MEGE ent, TE1 = 1.448 msec, and ΔTE = 0.636 msec. Each protocol acquisition protocols with different TE coverage to cover a ∗ had a scan time of less than 16 seconds, with the longest TEs wide range of T , is an extension of Wood’s work. It not ∗ being 27 msec and 11 msec, respectively. The other imaging only measures wide range of clinically relevant T , but also parameters, including matrix size (256 × 192), field of view allows us to simultaneously measure HIC and fat fraction for (38 cm), flip angle (25 ), receiver bandwidth (125 kHz), slice patients with hepatic steatosis. thickness (10 mm), and averages (2), were kept the same in all protocols. Among the two acquisitions, only one of the T fitted images with the higher correlation coefficient R of the 2. Materials and Methods fit, was selected to generate HIC map for clinical use. 2.1. Patient Population. This retrospective study was per- To validate the reproducibility of our method and test formed with the approval of our institutional review board, the stability of the system, a FerriScan phantom was used to measure T values with the same protocol for patients which waived the requirement for informed consent. We identified 28 consecutive patients with high ferritin who in every three to four months over a period of one year. underwent hematopoietic stem cell transplant at our insti- This phantom consists of 15 vials of aqueous manganese tution. Of these 28 patients, 19 had acute myelogenous chloride (MnCl ) solutions (0 mM to 3.2 mM) in 10 mM leukemia (AML), 3 had myelodysplastic disease (MD), 3 hydrochloric acid, which provide the T values from 3.0 msec to 40.0 msec, similar to the range of liver T values had chronic myelogenous leukemia (CML), and 3 had myelofibrosis. Clinically, it is believed that the amount of for patients. It is the same type of phantom used by Pierre et ferritin in blood reflects the amount of iron stored in the liver al. [7], who validated their technique for HIC measurement using the phantom and biopsy data. Their single echo SE [13]. Therefore, the serum ferritin level was measured within one month of the MRI scan in all patients. based technique would take 20–30 minutes to scan. Radiology Research and Practice 3 ∗ ∗ 2.3. Postprocessing MR Images . The T , R , and hepatic iron 3. Results 2 2 maps were processed off-line using an in-house developed For all 28 patients, the average T values ranged from 0.56± software tool based on MATLAB (Mathworks, Inc., Natick, 0.13 msec to 25.0 ± 2.1 msec. The corresponding HIC values MA, USA). The T curve was fitted to a monoexponential calculated from (1)–(3)werefrom1.2 ± 0.10 mg/g to 45.0 ± decay with three parameters using the Levenberg-Marquardt 10.0 mg/g dry weight. nonlinear least-squares method: Figure 1 shows a representative fitted signal versus time curves with long T value. Dual data acquisition method S(t) = A exp − + C,(1) was applied to the same patient, with fitted curves as shown in Figures 1(a) and 1(b),respectively. The T decay curve where A is the signal amplitude at time 0, with its initial fit acquired from the long TE protocol resulted in the better value set to 1.5 or 5.0 times of the maximum signal of the fit for the long T case, as demonstrated in Figure 1(a). ∗ ∗ ∗ first echo for each pixel for long T or short T cases; T is Therefore, the image from the long TE protocol was selected 2 2 2 initially set to 10 msec; and C represents noise whose initial to generate an iron map. Figure 2 is an example of a very fit value was set to the background noise. The background short T case with dual acquisition method in (a) and (b), noise was measured from a small ROI in the frequency respectively. The data acquired by the short TE protocol (for direction; its mean was used as a threshold to roughly mask short T values) provided better fit (b), as demonstrated by the image out before fitting. When there is fat in the liver, the its correlation coefficient and chi-square value. Therefore, in MRI signal in the liver can be represented by this case, images acquired by the short TE protocol were used to generate an iron map. In general, the more points sampled near the part of the curve with larger curvature, the more S(t) = W + F exp(iωt) exp − + C,(2) T reliable the generated results. For this reason, the longer TE and ΔTE spans were used for the long T fit, and the shorter where W is the water signal amplitude, F is the fat signal TE and ΔTE spans were used for the short T fit. As shown amplitude, and ω is the water-fat frequency difference that ∗ in Figure 2, the short T had faster decay and generated very is about 220 Hz at 1.5 T. Other parameters are the same as in low-intensity liver images. The signal intensity after the first (1). echo acquired by the long TE protocol, or after the third After fitting T curve using (1)or(2), the iron value, Fe echo acquired by the short TE protocol, was very close to the in mg/g (dry weight), was calculated based on a previously background noise and thus was much weaker than the signal validated linear relationship between iron and R value in from the long T (Figure 1)case. Hz [8]: Figure 3 displays a typical HIC map and its histogram of a 31-year-old man with acute myelogenous leukemia who F = 0.0254 · R +0.202. e (3) had very high levels of iron deposition. Figure 3(a) is the HIC map of the segmented liver, and the corresponding his- Both region of interest (ROI) and pixelwise fitting were togram with normal distribution fit is shown in Figure 3(b). implemented for a quick evaluation and T map generation. A combination of the HIC map and liver-only histogram A small ROI was drawn to plot the signal intensity versus provide local and global measurements of the response of the time curve for a quick review. If there was no signal liver to the treatment. oscillation, that means there was no fat in the liver, (1) Figure 4 is an example of T decay curves from a was used for curve fit; otherwise (2) was used for curve fit. hepatic steatosis patient acquired by our dual acquisition To reduce noise, the original images were smoothed by a ∗ method. There was no water-fat signal oscillation from 3 × 3 window kernel prior to curve fitting for short T the long TE acquisition protocol in Figure 4(a).However, cases. Since image intensity is not uniformly distributed due water-fat peaks were clearly demonstrated with the data to liver heterogeneity, tissue susceptibility, and surface coil acquired by the short TE protocol, as shown in Figure 4(b). sensitivity, local ROI measurements cannot represent the Figure 5 demonstrates that the long TE acquisition protocol global iron overload of the liver. After the pixelwise T curve is necessary to measure mild iron concentration as well as fitting, we segmented out the liver from the entire image PDFF for a hepatic steatosis patient. to form liver-only HIC map and generated a histogram of Shown in Figure 6 are the T measurement results (b) the liver-only HIC map with a normal distribution fit. This from a FerriScan Phantom consisting of fifteen different histogram shape, mean and standard deviations were used liquid tubes (a) with different T values. Three data sets were for longitudinal HIC comparison for patient followups. acquired with the same protocol for patient scan in a year The iron map generated by (3) was saved as a Digital with 3-4 months apart. Imaging and Communications in Medicine (DICOM) Figure 7 presents the relationship between R and ferritin image, its DICOM header was derived from original MRI for all patients. A paired Student’s t test was run between R image with different description and series number, and and ferritin values, with a correlation of 0.83 and P = 0.0001. was transferred to the institutional electronic medical record system, ClinicStation (Datalign, Inc., Houston, TX, USA), 4. Discussion for measurement and interpretation by radiologists. Radiol- ogists can obtain HIC values directly from the iron map by We have developed a simple and practical MRI strategy drawing an ROI. that provides the absolute HIC. By combining two different 4 Radiology Research and Practice Long TE protocol Long TE protocol ∗ 24 T = 23.16 ± 4.43 ms 140 2 R = 0.9963, χ = 8.561 ∗ T = 0.71 ± 0.66 ms R = 0.974, χ = 0.5356 60 14 0 510 15 20 25 30 0 5 10 15 20 25 30 Time (ms) Time (ms) Original Original 3-parameter fit 3-parameter fit (a) (a) Short TE protocol Short TE protocol T = 16.91 ± 22.9 ms R = 0.9514, χ = 11.0521 T = 0.57 ± 0.03 ms R = 0.9966, χ = 0.0855 90 18 02468 10 12 02468 10 12 Time (ms) Time (ms) Original Original 3-parameter fit 3-parameter fit (b) (b) ∗ ∗ Figure 2: Results of T curve fit from a patient with short T 2 2 ∗ ∗ Figure 1: Results of T curve fit from a patient with long T values 2 2 values by dual acquisition method (Figure 1(a)). Data was acquired by dual acquisition method. (a). Data was acquired by the long TE by the long TE protocol (Figure 1(b)). The same patient data was protocol. (b) the same patient data was acquired by the short TE acquired by the short TE protocol, there were more data points protocol. It is obvious for this case that the T fit error is increasing in the curvature area. It is obvious that data acquired by short TE with shorter TE coverage. The long TE data acquisition is necessary ∗ protocol generated better fit for short T case. for mild iron load patient. MEGE data acquisition protocols and an in-house image ΔTE and longer TE coverage data acquisition method should postprocessing tool, we cannot only measure a wide range of be used. Inadequate data acquisition protocol would result clinically relevant iron loads, from 1.2 mg/g to 45 mg/g (dry in great uncertainties in the value of HIC, as demonstrated weight), but also quantify water and fat simultaneously with in Figures 1 and 2. In this study, because the MRI scans confidence. This technique is now routinely used to evaluate were performed without prior knowledge of each patient’s HIC in all patients with suspected iron deposition disease in HIC, dual acquisition method covering different ranges of our institution. TE span was used. Since each acquisition took less than For patients with high levels of hepatic iron deposition 16 seconds, we were able to scan each patient with both (e.g., HIC > 15 mg/g), the T value was very short, and thus protocols (for different ΔTEs covering the entire range of a short ΔTE data acquisition method should be used. For possible T values) in two breath-holds. The data acquired patients with low levels of hepatic iron deposition (e.g., with the longer TE span generated better fits for the long T HIC < 3 mg/g), the T value was long and thus a relative long curve (Figure 1(a)), and the data acquired with the shorter Intensity Intensity Intensity Intensity Radiology Research and Practice 5 Long TE protocol T = 3.37 ± 0.36 ms R = 0.99, χ = 2.5891 0 5 10 15 20 25 30 Time (ms) (a) Original 3-parameter fit HIC histogram (a) HIC = 30.2 ± 5.3 mg/g 50 Short TE protocol Water = 71.31, fat = 42.88 35 T = 2.7 ± 0.11 ms R = 0.9982, χ = 0.2811 0 1020304050 HIC (mg/g) 02468 10 12 (b) Time (ms) Figure 3: A 31-year-old man with acute myelogenous leukemia, Original sixteen months after bone marrow transplantation. (a) is the HIC 3-parameter fit map of the segmented liver. (b) is its corresponding histogram with (b) normal distribution fit. This patient had very high iron deposition with mean HIC value of 30.20 mg/g (T < 1.0msec). Figure 4: An example of T curve fit from a liver steatosis patient with short T values and by dual acquisition method. (a) Data was acquired by the long TE protocol. The water and fat peaks were totally missed because data was sampled too slowly. (b) The same patient data was acquired by the short TE protocol with faster TE span generated better fits for the short T curves sample (shorter ΔTE). Water and fat peaks were caught and data (Figure 2(b)). was fitted with water and fat components. Again data acquired by A popular method to quantify liver iron by MR in the short TE protocol generated better fit for short T case. Europe is based on signal ratio of liver over reference tissue [11–18]fromdifferent images. Recently, this method has been further developed into a web-based iron measurement (URennes) [14] where proton-density-, T -, T -, and T - was 61.4% with a tendency of overestimate overload. They 1 2 2 weighted images have to be acquired based on their protocol, also found that the iron concentration by this method was and certain ROIs have to be measured in the liver and less reliable from 60 μmol/g to 170 μmol/g, corresponding to reference tissues as the inputs to calculate an adequate 3.35 mg/g to 9.5 mg/g. Another limitation of this method was iron value. Gandon et al. [16] also validated this technique that it could only measure iron up to 350 μmol/g (19.5 mg/g). with biopsy data on 1.5T system, declaring that it could This method was developed early in late 1990s [12] when be used on various MR scanners. Most recently, Castiella MEGE sequence was not commercially available. et al. [18] reevaluated this method using the data from Wood et al. [8] studied more than 100 patients for multiple institutions with different scanners in the period of HICs by T (SE), T (GRE) method. Among them, 21 were 2 2 1999 to 2006. They investigated the accuracy of this method liver biopsied and validated. They reported that the GRE with liver biopsy and found that the diagnostic accuracy sequence was able to adequately measure T values in the Pixels Intensity Intensity 6 Radiology Research and Practice Long TE protocol Water = 126.54, fat = 8.9 90 ∗ T = 20.64 ± 3.05 ms R = 0.9981, χ = 2.66 0 5 10 15 20 25 30 Time (ms) Original (a) 3-parameter fit Figure 5: An example of T decay curve fit from a liver steatosis patient with less iron deposition. The long TE protocol was adequate to catch the water and fat peak while minimum fit error for long T decay was achieved. entire range of iron overload. This method has recently been further validated by other investigators [9]. In our study, although the minimum TE was 1.448 msec, we were stillabletomeasure T values less than 1.0 msec from 8- echo, double-shot interleaved sequences, as demonstrated in Figure 2(b), where the SNR dropped to the background level at the fourth echo. This T value corresponds to a HIC value over 40 mg/g for that specific ROI, which is extremely S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13 S14 S15 high for a patient with acute myeloid leukemia [7, 8]. For Sample such high iron overload, raw images need to be smoothed to ∗ Scan 1 remove some noises. It is known that the T value for normal Scan 2 liver is ∼24 msec [20–22]. Therefore, we have demonstrated Scan 3 that our technique is able to measure the entire range of (b) possible clinical HIC values using the clinically available MRI sequence. With the maximum receiver bandwidth used Figure 6: Results of phantom validation. (a) is the image of in our protocol, the first TE could be reduced further by the FerriScan phantom consisting of fifteen different tubes with using a 128 × 128 or even 64 × 64 acquisition matrix. This different solutions (different T values). The sample numbers are might be beneficial for SNR for cases with very short T defined on the left. (b) is the T measurement from three scans in a 2 2 values (<1.0 msec). Meanwhile, Figure 1 demonstrates the year using the same clinical protocols. Consistent measurement was importance of long TE acquisition for long T decay case. demonstrated in this plot. Short TE acquisition for long T decay case would result in larger measurement uncertainty, a limitation of Wood’s method. in Figure 2(b), we did observe significant noise from data The slope and intercept for the linear equation (3) acquisition, especially close to the high tissue susceptibility between HIC and R under 1.5-T was determined by Wood area and low coil sensitivity area. For patients with high et al. [8] using biopsy results. It was also calibrated and HIC values, having an HIC map is necessary to view the confirmed with biopsy values by Hankins et al. [9]. In our image nonuniformity, an example is shown in Figure 3(a). study, we used Wood et al.’s results and did not validate However, this HIC image map is not enough to track our HIC values against those from biopsies. In addition, the overall mean HIC value and standard deviation. A Wood pointed out that their calibration was validated from liver-only HIC histogram with normal distribution fit, as 1.3 mg/g to 32.9 mg/g dry weight. There was an outlier shown in Figure 3(b), can provide the mean and standard with HIC value of 57.8 mg/g, it was removed from their deviations of the liver’s iron concentration. We have found statistic calibration. With the minimum TE of 1.448 msec that a combination of HIC image map and its histogram in our dual acquisition method, although we can measure distribution is extremely useful for patient followup after and fit the T value as short as 0.57 msec, as demonstrated treatment. Intensity T (ms) 2 Radiology Research and Practice 7 R versus ferritin observed curve fit improvement. Therefore, in our study, we assumed that the T decay was identical for water and fat signal (2). The independent T decay model needs to be further studied with a larger number of patients. Researchers [21, 22, 30, 31] have demonstrated that iron measurement from T method has high reproducibility and inter-MRI scanner agreement. Using FerriScan phantom, we have confirmed the reproducibility (Figure 6)of T measurements for all samples. The average error was 2.1% with the maximum error of 3.9%. The largest error mainly came from nonuniform image due to phantom/slice position between each scans, seen in sample 3 and 6. The consistent measurement from fixed phantom samples further demon- strated that our technique is quite promising for patient followups. In our study, we have found that consistency in data acquisition and analysis is vital for patient followups. As mentioned earlier, there was a reasonably good corre- lation between the R and ferritin values (Figure 7)inour study. Although the serum ferritin values have been used as a 0 5000 10000 15000 marker for body iron amount [19], some researchers [9, 10] Ferritin (ng/ml) have shown that the ferritin values do not necessarily reflect y = 127.97 + 0.10375x, R = 0.82969 the actual total body iron burden. Therefore, using serum R (Hz) ferritin to represent iron load is still a debatable topic. On ∗ the other hand, HIC as measured by an MEGE sequence Figure 7: Scattergram of R values calculated from MEGE MRI does represent the actual hepatic iron burden, as supported acquisition versus serum ferritin values. A good correlation between by several biopsy reports [7–9]. Using the dual acquisition R and ferritin is clearly demonstrated. method described in this study, the MEGE sequence can be used in the clinical decision-making with higher confidence. Our dual acquisition method cannot only cover wide Image quality will affect the iron measurements. First, range of clinical relevant T decay (all iron deposition), but susceptibility artifacts will affect our measurement of HIC. also quantify iron accurately for the patient with superim- Shimming could help to reduce some B0 inhomogeneity, and posed steatosis, as demonstrated in Figure 4. For hepatic positioning the patient carefully so that liver is not close to steatosis patient with less iron deposition, the long TE the edge of coil will help to generate more uniform image. protocol can detect water and fat oscillation, as shown in Since the whole HIC map is generated, radiologist can always Figure 5. Meanwhile, the short TE protocol with shorter ΔTE put an ROI to the most homogeneous part to measure HIC value. The histogram from liver-only HIC provides overall can detect the water and fat oscillation for patient with high iron deposition.Itiswellknown thatwater andfat will be mean and standard deviations, a useful distribution to mon- in phase at time 4.6 msec and out of phase at 2.3 msec on itor treatment. Meloni et al. [25] have studied single-slice versus multislice HIC distribution of HIC by T method and 1.5 T system. By setting the echo time at the multiple time of 2.3 msec, Sirlin and Reeder [26–28] studied fat concluded that the single-slice measurement is enough in the quantification by measuring PDFF using an MEGE sequence. clinical application as long as major vessels are avoided. Sec- Their methods can quantify fat and water, as well as generate ond, patient’s motion affects the T measurement, breathing R map with four to eight echoes. Their main purpose was resulting in severe nonuniform artifact in the liver image. fat assessment; therefore, the minimum TE and ΔTE were Breath instruction should be delivered clearly before data not short enough to quantify high iron overload cases. For acquisition and patients need to be trained to hold breath. extremely high iron overload such as T < 1.0 msec, steatosis Our technique has some limitations. One limitation is the lack of validation of patients with combined iron and could not be detected by image technique because water and fat peak may not be MRI visible. In our study, no fatty liver fat deposition in the liver, especially in high HIC (T < was detected with T value less than 2.0 msec. To improve 2.0 msec) and high PDFF cases. The second limitation is that the TEs are not set at the multiple of 2.3 msec for long the accuracy of HIC and fat measurement, a better pulse sequence is needed with the minimum TE and ΔTE less than TE data acquisition protocol. This could result in PDFF 1.0 msec, the maximum TE greater than 20.0 msec, and some underestimation for minor iron load patient, such as the case of the echo should be set at multiple times of 2.3 msec. in Figure 5. Another minor problem is that the multiecho Chebrolu et al. [29] pointed out that the independent T -weighted images have to be transferred from the MR scanner to a local workstation for postprocessing, and it T decay for water and fat model may improve accuracy of water-fat quantification using a 3D MEGE sequence, takes several minutes for an HIC map to be generated. Since particularly for high fat fraction and short T case. We have the postprocessing on a local workstation is automatic, it is still practical. It would be ideal to generate the HIC map used independent T model to fit the signal decay curve. For our limited number of steatosis patients, we have not immediately after image acquisition on the MRI scanner. R (Hz) 2 8 Radiology Research and Practice 5. Conclusion Roentgenology, vol. 168, no. 5, pp. 1205–1208, 1997. [13] M. Queiroz-Andrade, R. Blasbalg, C. D. Ortega et al., “MR We have developed a simple, quick, and reliable method that imaging findings of iron overload,” Radiographics, vol. 29, no. can noninvasively measure hepatic iron overload using a 6, pp. 1575–1589, 2009. [14] Y. Gandon, “Iron, liver and MRI,” University of Rennes, http:// standard 1.5-T MRI scanner. By combining different MEGE www.radio.univ-rennes1.fr/Sources/EN/Hemo.html. data acquisition strategies and an in-house image postpro- [15] J. M. Alustiza, ´ J. Artetxe, A. Castiella et al., “MR quantification cessing tool, our technique cannot only measure a wide of hepatic iron concentration,” Radiology, vol. 230, no. 2, pp. range of clinically relevant HIC values from 1.2 mg/g to 479–484, 2004. greater than 45.0 mg/g dry weight, but also quantify fat/water [16] Y. Gandon, D. Olivie, ´ D. Guyader et al., “Non-invasive assess- fraction simultaneously. It generates an absolute HIC map ment of hepatic iron stores by MRI,” The Lancet, vol. 363, no. and transfers it to the electronic medical record system, 9406, pp. 357–362, 2004. which makes it convenient for the physicians to diagnose and [17] A. W. Olthof,P.E.Sijens, H. G. Kreeftenberg,P.Kappert,E. measure hepatic overload. J. van der Jagt, and M. Oudkerk, “Non-invasive liver iron concentration measurement by MRI: comparison of two validated protocols,” European Journal of Radiology, vol. 71, References no. 1, pp. 116–121, 2009. [18] A. Castiella, J. M. Alustiza, ´ J. I. Emparanza, E. M. Zapata, B. [1] T. Robak and A. Wierzbowska, “Current and emerging ther- apies for acute myeloid leukemia,” Clinical Therapeutics, vol. Costero, and M. I. D´ıez, “Liver iron concentration quantifi- cation by MRI: are recommended protocols accurate enough 31, no. 2, pp. 2349–2370, 2009. [2] J. J. Cornelissen, W. L. J. Van Putten, L. F. Verdonck et al., for clinical practice?” European Radiology,vol. 21, no.1,pp. 137–141, 2011. “Results of a HOVON/SAKK donor versus no-donor analysis of myeloablative HLA-identical sibling stem cell transplanta- [19] A. W. Olthof, P. E. Sijens, H. G. Kreeftenberg et al., “Correla- tion between serum ferritin levels and liver iron concentration tion in first remission acute myeloid leukemia in young and middle-aged adults: benefits for whom?” Blood, vol. 109, no. determined by MR imaging: impact of hematologic disease 9, pp. 3658–3666, 2007. and inflammation,” Magnetic Resonance Imaging, vol. 25, no. 2, pp. 228–231, 2007. [3] D. Vicente, T. Lamparelli, F. Gualandi et al., “Improved out- come in young adults with de novo acute myeloid leukemia in [20] H. Chandarana, R. P. Lim, J. H. Jensen et al., “Hepatic iron deposition in patients with liver disease: preliminary expe- first remission, undergoing an allogeneic bone marrow trans- plant,” Bone Marrow Transplantation, vol. 40, no. 4, pp. 349– rience with breath-hold multiecho T2 -weighted sequence,” American Journal of Roentgenology, vol. 193, no. 5, pp. 1261– 354, 2007. [4] D.D.Stark,N.M.Bass, andA.A.Moss, “Nuclear magnetic 1267, 2009. [21] M. Westwood, L. J. Anderson, D. N. Firmin et al., “A single resonance imaging of experimentally induced liver disease,” Radiology, vol. 148, no. 3, pp. 743–751, 1983. breath-hold multiecho T2 cardiovascular magnetic reso- nance technique for diagnosis of myocardial iron overload,” [5] R. Engelhardt, “Liver iron quantification: studies in aqueous iron solutions, iron overloaded rats, and patients with heredi- Journal of Magnetic Resonance Imaging, vol. 18, no. 1, pp. 33– 39, 2003. tary hemochromatosis,” Magnetic Resonance Imaging, vol. 12, no. 7, pp. 999–1007, 1994. [22] L. J. Anderson, S. Holden, B. Davis et al., “Cardiovascular ∗ ∗ [6] Z.J.Wang, J. C. Haselgrove,M.B.Martinetal., “Evaluationof T2 (T2 ) magnetic resonance for the early diagnosis of myocardial iron overload,” European Heart Journal, vol. 22, no. iron overload by single voxel MRS measurement of liver T2,” Journal of Magnetic Resonance Imaging, vol. 15, no. 4, pp. 395– 23, pp. 2171–2179, 2001. [23] S. D. Chirnomas, M. Geukes-Foppen, K. Barry et al., “Practical 400, 2002. [7] T. G. St Pierre, P. R. Clark, W. Chua-Anusorn et al., “Nonin- implications of liver and heart iron load assessment by T2 -MRI in children and adults with transfusion-dependent vasive measurement and imaging of liver iron concentrations using proton magnetic resonance,” Blood, vol. 105, no. 2, pp. anemias,” American Journal of Hematology, vol. 83, no. 10, pp. 781–783, 2008. 855–861, 2005. [8] J. C. Wood, C. Enriquez, N. Ghugre et al., “MRI R2 and [24] V. Positano, B. Salani, A. Pepe et al., “Improved T2 assess- ment in liver iron overload by magnetic resonance imaging,” R2 mapping accurately estimates hepatic iron concentration in transfusion-dependent thalassemia and sickle cell disease Magnetic Resonance Imaging, vol. 27, no. 2, pp. 188–197, 2009. [25] A. Meloni, A. Luciani, V. Positano et al., “Single region of patients,” Blood, vol. 106, no. 4, pp. 1460–1465, 2005. ∗ ∗ [9] J.S.Hankins,M.B.McCarville, R. B. Loeffler et al., “R2 interest versus multislice T2 MRI approach for the quantifi- magnetic resonance imaging of the liver in patients with iron cation of hepatic iron overload,” Journal of Magnetic Resonance Imaging, vol. 33, no. 2, pp. 348–355, 2011. overload,” Blood, vol. 113, no. 20, pp. 4853–4855, 2009. [10] A. W. K. Leung,W.C.W.Chu,W.W.M.Lam,V.Lee,and C. [26] H. Yu, C. A. McKenzie, A. Shimakawa et al., “Multiecho recon- struction for simultaneous water-fat decomposition and T2 K. Li, “Magnetic resonance imaging assessment of cardiac and liver iron load in transfusion dependent patients,” Pediatric estimation,” Journal of Magnetic Resonance Imaging, vol. 26, no. 4, pp. 1153–1161, 2007. Blood and Cancer, vol. 53, no. 6, pp. 1054–1059, 2009. [11] Y. Gandon, D. Guyader, J. F. Heautot et al., “Hemochromato- [27] M. Bydder,T.Yokoo,G.Hamiltonetal., “Relaxationeffects in the quantification of fat using gradient echo imaging,” sis: diagnosis and quantification of liver iron with gradient- echo MR imaging,” Radiology, vol. 193, no. 2, pp. 533–538, Magnetic Resonance Imaging, vol. 26, no. 3, pp. 347–359, 2008. [28] S. B. Reeder, I. Cruite, G. Hamilton, and C. B. Sirlin, “Quanti- [12] O. Ernst, G. Sergent, P. Bonvarlet, V. Canva-Delcambre, J. C. tative assessment of liver fat with magnetic resonance imaging ´ and spectroscopy,” Journal of Magnetic Resonance Imaging, vol. Paris, and C. L’Hermine, “Hepatic iron overload: diagnosis and quantification with MR imaging,” American Journal of 34, no. 4, 2011, spcone. Radiology Research and Practice 9 [29] V. V. Chebrolu, C. D. G. Hines, H. Yu et al., “Independent estimation of T 2 for water and fat for improved accuracy of fat quantification,” Magnetic Resonance in Medicine, vol. 63, no. 4, pp. 849–857, 2010. [30] M. A. Westwood, D. N. Firmin, M. Gildo et al., “Intercentre reproducibility of magnetic resonance T2 measurements of myocardial iron in thalassaemia,” International Journal of Cardiovascular Imaging, vol. 21, no. 5, pp. 531–538, 2005. [31] M. A. Westwood, L. J. Anderson, D. N. Firmin et al., “Inter- scanner reproducibility of cardiovascular magnetic resonance T2 measurements of tissue iron in Thalassemia,” Journal of Magnetic Resonance Imaging, vol. 18, no. 5, pp. 616–620, 2003. MEDIATORS of INFLAMMATION The Scientific Gastroenterology Journal of World Journal Research and Practice Diabetes Research Disease Markers Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 International Journal of Journal of Immunology Research Endocrinology Hindawi Publishing Corporation Hindawi Publishing Corporation http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 Submit your manuscripts at http://www.hindawi.com BioMed PPAR Research Research International Hindawi Publishing Corporation Hindawi Publishing Corporation http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 Journal of Obesity Evidence-Based Journal of Journal of Stem Cells Complementary and Ophthalmology International Alternative Medicine Oncology Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 Parkinson’s Disease Computational and Behavioural Mathematical Methods AIDS Oxidative Medicine and in Medicine Research and Treatment Cellular Longevity Neurology Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

Journal

Radiology Research and PracticeHindawi Publishing Corporation

Published: Dec 11, 2012

References