Magnetic resonance imaging quantification of liver iron

Abstract

Iron overload is the histologic hallmark of hereditary hemochromatosis and transfusional hemosiderosis but also may occur in chronic hepatopathies. This article provides an overview of iron deposition and diseases where liver iron overload is clinically relevant. Next, this article reviews why quantitative noninvasive biomarkers of liver iron would be beneficial. Finally, we describe current state-of-the-art methods for quantifying iron with MR imaging and review remaining challenges and unsolved problems.

Figure 1

Conventional T2*-weighted gradient echo and T2w-weighted imaging is a well-established qualitative method for detecting iron overload within the liver and assessing involvement of other organs. Shown are breath-held gradient echo images with echo times (TEs) of 2.3, 4.6, and 9 ms and a respiratory-triggered fat-saturated fast spin echo image with TE of 70 ms at 1.5T through the same slice of the liver in a male patient with transfusional iron overload. Notice abnormally low signal intensity of liver, spleen, and marrow on all images, indicating iron overload in these recticulo-endothelial tissues. The pancreas, which has normal signal intensity, is spared. The pattern of organ involvement is typical for transfusional overload. The liver is mildly hypointense at 2.3 but then shows progressive signal loss with increasing TE on the gradient echo images, indicating fast T2* decay. The spleen is markedly hypointense at 2.3 ms, indicating even faster T2* decay. As its signal intensity at 2.3 ms is already very low, the spleen does not appreciably lose additional signal with increasing TE.

Figure 2

T2*-weighted gradient echo (GRE) sequences are more sensitive to the presence and distribution of hepatic iron overload than T2-weighted single shot fast spin echo (SSFSE) sequences. Shown are co-localized SSFSE (top row) and GRE (bottom row) images in a woman with transfusional iron overload at baseline and at 4 and 8 months after chelation therapy. The T2-weighted SSFSE images look similar at all three time points, but the T2*-weighted GRE images show diffuse iron overload at baseline, partial regression of iron overload at 4 months, and near complete regression at 8 months. Notice that at 4 months, the GRE image shows normal signal intensity in a branching peri-portal pattern with persistent hypointensity in intervening parenchyma and along the periphery, suggesting that different parts of the liver may clear excess iron at different rates.

Figure 3

In patients with pre-existing cirrhosis, iron may accumulate in regenerating nodules. Iron-laden regenerating nodules are known as siderotic nodules. Shown are T2*-weighted gradient echo images after administration of a gadolinium based contrast agent in two patients with cirrhosis secondary to hepatitis C viral infection. Siderotic nodules are hypointense due to T2* shortening effects of iron. In the patient on the left, scattered siderotic nodules are evident in a patchy distribution, while in the patient on the right they are diffusely distributed. Hyperintense reticulations in the liver in both patients represent gadolinium-enhanced fibrotic bands. These are more conspicuous in the patient on the right due to greater contrast between the hyperintense fibrotic bands and the diffusely distributed hypointense siderotic nodules.

Figure 4

Patients with thalassemia and secondary iron overload have preferential involvement of reticuloendothelial tissues. Shown are dual-echo gradient echo images obtained at 3T with echo times (TEs) of 2.3 and 5.8 ms. Notice low signal of liver and spleen at 2.3 ms and incremental signal loss at 5.8 msec due to iron-mediated T2* shortening. In principle, the signal loss of the liver at 5.8 ms (nominally out of phase at 3T) could be attributed to steatosis rather than iron overload, which is a limitation of the IP-OP sequence design commonly implemented at 3T. In this case, a fat quantification sequence (not shown) excluded the presence of concomitant fat. The pancreas and kidney are spared, suggesting the storage capacity of the reticuloendothelial system has not been exceeded.

Figure 5

Patients with SCD and secondary iron overload may have iron accumulation in the renal cortex. In this patient with SCD and history of transfusion therapy, notice low signal in the renal cortex and liver on T1-weighted dual-echo gradient echo images acquired at 1.5T with echo times as shown. The low signal is more pronounced on the second echo (TE = 4.6 ms). As illustrated in this case, dual-echo T1-weighted imaging can be used to detect iron overload in tissue. Signal loss on the second echo compared to the first echo indicates short T2* decay and suggests the presence of iron. As discussed in the text, hepatic iron overload does not occur in patients with SCD in the absence of transfusion therapy.

Figure 6

Quantification of iron can be performed using multiple gradient echo images acquired with T₁ weighting (TE=4ms, flip=90°), and increasing amounts of T₂* weighted (TE=4, 9, 14 and 21ms, with flip = 20°) according to Gandon et al. At least three ROI's are placed in the liver, and two in the muscle to provide normalization for B1 sensitivity and provide signal ratios. ROI's are propagated to all 5 images and values entered in the website available on-line at: http://www.radio.univ-rennes1.fr/Sources/EN/HemoCalc15.html. Based on the signal intensities entered, an estimated LIC is provided automatically.

Figure 7

R₂* mapping can be performed in a single breath-hold using rapid gradient echo methods that acquire multiple images at increase echo times, within the same TR. Results are typically displayed as an R₂* map (bottom right), where areas of high iron concentration appear bright. Alternatively, results can also be displayed as a T₂* map (bottom left), which may be more intuitive since regions of elevated iron appear dark, corresponding to the appearance of iron overloaded tissue in heavily T₂* weighted images. Both approaches are equally valid.

Figure 8

R2* map (below) is the reciprocal of T2* map (above). Both maps depict the distribution of iron in the liver. In this patient, the left lobe has greater iron content than the right lobe, and it appears darker than the right lobe on the T2* map and brighter on the R2* map. As illustrated in this case, R2* maps may depict heterogeneity in iron distribution to better advantage than T2* maps because of their wider gray-level dynamic range.

Figure 9

R2* map shows scattered siderotic nodules (arrows) in a patient with alcoholic cirrhosis. C = cava. Ao = aorta. St = stomach. Sp = spleen. The liver margins have been highlighted in black to improve demarcation of the liver.

Figure 10

R₂* maps can be used to monitor treatment for iron overload, as with this 52 year old female with known genetic hemochromatosis. R₂* maps acquired 1 year after treatment with phlebotomy demonstrate a marked decrease in R₂* from 185s^-1 to 96s^-1, with a corresponding decrease in the serum ferritin, which is an indirect marker of iron overload. These particular R₂* maps were acquired with the Yu method which automatically produces an R₂* map as part of the correction needed for fat quantification.

Figure 11

Fat corrupts the ability of gradient echo methods to quantify T₂* using conventional multi-echo imaging methods. In the simulation shown in this figure, the signal from a water-fat mixture with increasing amounts of fat at 0% (black), 20% (red) and 40% (blue) are shown for a true T₂*=25ms at 1.5T. Even when images are acquired “in-phase” (asterisks) the fitted values of T₂* are inaccurate when fat is present. This occurs because the spectral complexity of fat causes fat to interfere with itself and accelerates the effective signal decay. In order to measure T₂* accurately, simultaneous measurement of fat and T₂* is necessary, including spectral modeling of fat, such as performed by the methods of Bydder et al and Yu et al. Without simultaneous measurement of fat and T₂* with spectral modeling, it is not possible to measure fat-concentration and iron concentration (indirectly through T₂*), when both iron and fat are present.

Figure 12

The number of echoes used to quantify iron (indirectly, through measurement of T₂*) also impacts the ability of MRI to measure T₂*, when fat is present. The presence of fat creates a complicated signal pattern due to the interference of multiple fat peaks, leading to signal decay that is no longer monoexponential. Without simultaneous measurement of fat and T₂*, including spectral modeling, the ability of MRI to measure T₂* accurately is corrupted, even when images are acquired at “in-phase” echo times (squares). In this simulation the true T₂*=25ms, and the fat-fraction is 40%.

Figure 13

IOP imaging in a 60 year old male with known genetic hemochromatosis shows paradoxical decrease in signal on the in-phase images (middle column) because these images are more heavily T₂* weighted (TE=4.6ms, compared to 2.3ms for opposed-phase images). R₂* maps measured using the Yu method demonstrate dramatically shortened T₂* in the liver (4.0ms) as well as the pancreas (arrow, T₂*=5.0ms). Note that while the liver and pancreas are affected, the spleen is spared, typical of genetic hemochromatosis.

Figure 14

Fat and iron corrupt the ability to detect and quantify the other using IOP imaging. In this obese patient with hemosiderosis, the signal intensity of the in-phase and opposed-phase images are essentially equal, suggesting that the liver is normal. However, quantitative imaging with the complex method of Yu et al shows markedly decreased T₂* (9.6ms) and abnormally high fat concentration (13.5%) demonstrating how the presence of iron and fat corrupt the ability to measure the other, unless both are measured simultaneously. Also, note the shortened T₂* in the spleen, consistent with hemosiderosis. Images courtesy Vasanawala, MD, PhD, Stanford University, Stanford, CA and Huanzhou Yu, PhD, MR Global Applied Science Lab, GE Healthcare, Menlo Park, CA.

Figure 15

High fat concentrations and elevated iron can coexist, necessitating the use of methods such as those by Bydder et al and Yu et al that can simultaneously estimate PDFF and T₂*. In this obese patient with hemosiderosis, the PDFF measured with the complex method of Yu was 36.5%, and the T₂* was 9.4ms, both highly abnormal. Images courtesy Vasanawala, MD, PhD, Stanford University, Stanford, CA and Huanzhou Yu, PhD, MR Global Applied Science Lab, GE Healthcare, Menlo Park, CA.

Figure 16

Simultaneous measurement of fat-fraction and R₂* using the Bydder method. Multiple unusual hypodense lesions were identified at CT (not shown), concerning for metastatic disease. Conventional IOP imaging reveals multiple hypointense nodules, best seen on OP images (left) corresponding to the hypodense nodules seen on CT. Fat signal fraction (FSF) calculated with 2-point IOP imaging (equation 2) shows no evidence of liver fat, because of T₂* shortening. PDFF and R₂* maps measured with the Bydder method show diffuse fatty liver with nodular areas of focal fat, as well as elevated R₂* (shortened T₂*) in both the liver and spleen suggestive of hemosiderosis.