One of the biggest sources of failure for hepatic MR Elastography (MRE) is low magnitude image signal which leads to noisy wave data and uninterpretable elastograms. This failure occurs in approximately 4% of exams¹ and can be caused by issues such as iron overload or anemia. Specialized spin-echo-based acquisitions have been designed to increase the MR signal in patients with short T2*. In an earlier study, we demonstrated that these acquisitions do not result in a stiffness bias when applied to patients with normal liver signal.² This study investigates the ability of these new sequences to allow stiffness quantification in patients in which the standard GRE acquisition failed due to low MR signal.

This retrospective study used clinical data acquired during the period between March 2013 and September 2015 in accordance with our institutional review board. During this period, SE and SE-EPI data were acquired in addition to standard GRE MRE for all patients. For all clinical analysis, low-SNR areas are masked out from the elastogram analysis using an MRE inversion confidence threshold of 0.95. The automated tool ALEC,³ which is the standard method of MRE analysis at our institution, was used to determine cases in which GRE MRE had a low inversion confidence (<30% of the liver with confidence level >0.95) and the SE and SE-EPI MRE data were available. Cases with non-signal-related failure (e.g., driver disconnection or patient motion) were excluded, leaving a set of 48 exams. Liver segmentations were performed automatically, with manual modification when necessary, and GRE, SE, and SE-EPI images were compared in terms of the following:

1. SNR: average signal within the liver divided by the signal standard deviation in the liver.

2. Confidence level: average confidence level (between 0 and 1) of the MRE inversion within the liver.

3. Normalized high-confidence area (NHCA): number of liver voxels with >0.95 inversion confidence divided by the total number of voxels in the liver.

These parameters were also calculated in 130 successful GRE MRE images retrieved from a smaller overlapping time period for comparison. In cases where GRE MRE had low signal but both SE and SE-EPI were successful (NHCA>30%), liver stiffnesses were calculated from automated ROIs and compared between the spin-echo sequences using a paired equivalence t-test (using JMP9.0, Cary, NC) with a significance level of 0.05 and an equivalence margin of 5% of the measured stiffness. All acquisitions were performed at 1.5 T using 60 Hz motion.⁴

An example of the images obtained with the three sequences on a patient with low GRE MRE signal is shown in Figure 1. The spin-echo sequences provided sufficient signal to produce reliable liver stiffness measurements. As can be seen from Figure 2 and Figure 3, the SNR and confidence levels in patients with low liver signal are very low, while the specialized SE and SE-EPI acquisitions improve the signal to the level of successful GRE MRE exams. The SE and SE-EPI sequences salvaged 44/48 and 45/48 exams, respectively, by increasing NHCA to over 30%. Using SE, 3 cases continued to have low signal and 1 case had an unrelated driver problem. In SE-EPI, 2 continued to have low-signal and 1 case, different from the one which failed in SE, had a driver problem. The stiffnesses calculated from the SE and SE-EPI exams are shown in Figure 4. The mean and standard deviation of the stiffnesses were 3.39 ± 1.92 kPa and 3.46 ± 1.80 kPa, respectively, and were equivalent with a p<0.001 at a 5% equivalence margin. Stiffness values for these patients were reported clinically based primarily on the SE data. The causes of signal-related GRE MRE failure were not investigated as the majority of exams were effectively salvaged by the use of spin-echo sequences. The specialized spin-echo sequences provide an effective means of imaging patients with iron overload, metallic implant, or other conditions which lead to low liver signal, and may have a greater impact at 3T. Since GRE signal-related failure is relatively infrequent (<5% of cases), GRE is likely to remain the standard for 1.5T 2D MRE.This study has demonstrated that specialized spin-echo sequences enable MRE to be performed in cases where standard GRE MRE fails by bringing the SNR and confidence to the levels characteristic of successful GRE MRE exams. This complements our previous study which demonstrated that SE and SE-EPI sequences provide unbiased stiffness estimates compared to GRE MRE in exams with normal signal levels. This further supports the use of specialized spin-echo-based sequences to reduce the failure rate of MRE exams.