Physicians (radiologists, hepatologists), image analysts, and physicists with an interest in MR-based liver fibrosis quantification.Liver fibrosis is an important public health problem, with substantial morbidity and mortality due to progression to cirrhosis (the end stage) and hepatocellular carcinoma. Elastography methods are widely used for non-invasive assessment of liver fibrosis. Commercially available methods rely on wave generation through an external driver to produce shear waves.¹ These require additional hardware and are mostly valid for the right liver. Alternatively, internal deformation caused by cardiac-induced motion may be assessed by MRI tagging of the liver. The harmonic phase (HARP) method allows fast and automated analysis of tagging images to assess tissue displacement and strain at every point in the image.² The primary aim was to assess the diagnostic accuracy of MRI cine-tagging using inherent cardiac motion for the staging of liver fibrosis. The secondary aim was to compare different post-processing techniques.This substudy of the prospective Elastography Liver Fibrosis (ELF) trial was approved by the institutional review board. Adult patients were included if they underwent liver biopsy as part of their standard of care for suspected or known chronic liver disease caused by hepatitis B virus, hepatitis C virus or nonalcoholic steatohepatitis. All studies were performed on a 3.0T clinical MRI system (Achieva TX, Philips Healthcare, Best, The Netherlands). For each patient a 2D multi-slice gradient-echo sequence with tagging was acquired with peripheral pulse-wave triggering. Tagging was performed with the spatial modulation of magnetization (SPAMM) preparation sequence which creates a modulation of the underlying image by a sinusoidal magnetization pattern. The following parameters were used: repetition time (TR), 4.9 ms; echo time (TE), 2.8 ms; number of phases per cardiac cycle, 12-15; flip angle, 10°; field of view, 420 x 420 mm2; in-plane resolution, 1.3 mm x 1.3 mm; slice thickness, 8 mm; 16 mm gap; tag spacing, 8 mm; tag orientation, 0 and 90°; receiver bandwidth, 430 Hz/pixel; SENSE acceleration factor, 2; and number of averages, 1. Images were acquired with 4 consecutive breath holds at end expiration by study participants. Total acquisition time was approximately 16 s per slice and was adapted to patient cardiac frequency. A research assistant performed the image post-processing using publicly available software (HARP for MATLAB, John Hopkins University, Baltimore MD). The sinusoidal modulation of the images produces harmonic peaks in the Fourier domain, which provide both magnitude and phase information about the image. As the phase is constant for a material element of the inspected volume throughout time, it is possible to track the movements of each point using the corresponding HARP images.³ It is therefore possible to extract the strain tensor at every point in the image. Values for the principal strain were extracted from the resulting images in four different ways: an average over a region of interest (ROI) directly under the heart inside the liver, the same ROI averaged over two slices, the average over the 350 pixels with the highest strain values in the liver and the same value normalized by the strain in a small ROI inside the heart. The research assistant was blinded to the liver biopsy results. The correlation between strain values and liver fibrosis stages was assessed by Spearman's rho. The diagnostic accuracy for classification of liver fibrosis was assessed by the Kruskal-Wallis test and the Mann-Whitney U test was conducted for distinguishing between ≤F2 and ≥F3 stages for each strain assessment method.Eight patients underwent MRI cine-tagging and had available biopsy results for the assessment of fibrosis stage. Figure 1 shows representative magnetization grids with overlaid maximum strain maps in patients with a low fibrosis score and cirrhosis. The Spearman correlation coefficients (ρ) between fibrosis stage and ROI mean, ROI averaged over two slices, maximum principal strain over the liver and maximum normalized were -0.678, -0.866, -0.913, and -0.861, respectively. Figure 2 shows the observed decrease in strain values in higher fibrosis stages and the associated Mann-Whitney p-values for the detection of ≥F3 liver fibrosis. These were significant (p<0.05) for three of the methods, excluding the ROI mean value method. The Kruskal-Wallis p-value for fibrosis quantification by strain measurement approached significance for the maximum strain over the whole liver method, at 0.054.This study shows that maximum strain in the liver decreases with an increasing fibrosis stage. The strong correlation between maximum strain and fibrosis stage suggests that this method may be used for non-invasive staging of liver fibrosis. Further analysis will be conducted on upcoming study subjects.