Introduction
mpMRI methods in clinical practice
-
T2/T2*: Since R2 = 1/T2 and R2* = 1/T2*, we will talk about them indiscriminately. Iron content can be measured using spin-density projection-assisted R2 [51,52] or T2* transverse relaxation, for example with GRE sequences [16,28,40,53]. These methods are standardised across scanners [42,51] and commercially available (Resonance Health, Australia and Perspectum, UK, respectively). Semi-automated post-processing services with same day turnarounds are now possible for T2*. Fibrosis, fat, and other hepatic cellular pathology contribute to R2 and R2* and interfere with liver iron content estimation [54‐56]. The effect of fat has accuracy implications in NAFLD [55] but appears to be relatively small and may be minimised by mathematical correction [57,58]. In, addition, R2* can be obtained simultaneously with PDFF with Dixon-based sequences [55,56,59]. The confounding effect of fibrosis may be overcome with newer processing methods [54].
-
Proton density fat fraction (PDFF): PDFF is a ratio, expressed as a percentage, of the fraction of the MRI-visible protons attributable to fat divided by all MRI-visible protons in that region of the liver attributable to fat and water. Taking advantage of the chemical shift between fat and water, pulse sequences can be used to acquire images at multiple echo times at which fat and water signals have different phases relative to each other [60, 61]. PDFF can be performed with very high precision using a multiple echo spoiled GRE sequence with > 3 echo times. To avoid biasing the PDFF measurement it is important to image with a low flip angle to minimise T1 weighting (such as flip angle 5°, TR = 12 ms at 1.5T,flip angle 3°, TR = 14 ms at 3T) [62]. A complete set of sequence recommendations has been formulated by the quantitative imaging biomarkers alliance group (QIBA) [63,64]. MRI-PDFF is easier to perform, has high reproducibility [65, 59, 64] and reflects fat distribution on at least one cross-sectional slice rather than a few voxels, so has replaced measurement of triglyceride content using 1H MR spectroscopy even in guideline recommendations [17, 66] including for diabetes [19]. To minimise T1 bias (fat has shorter T1 than water), a low flip angle is used, along with acquisition or algorithmic corrections for T2* effects [59,67‐69]. Developments in echo times [70‐72] and processing improve sensitivity to field inhomogeneities, signal-to-noise ratios and sensitivity at PDFF > 50% [73,74]. These also highlight that PDFF accuracy is not meaningfully confounded by any of age, sex, BMI, inflammation or fibrosis [75,76].
-
Magnetic resonance elastography (MRE): MRE uses low frequency mechanical shear waves to cause liver vibrations that are detected by MRI, based on a modified phase contrast pulse sequence [34, 77]. 3D-MRE takes advantage of additional spin-echo echo-planar-imaging (SE-EPI) to capture shear wave displacements along three dimensions, and images the entire liver rather than regions of interest (ROI) but to date is not FDA-cleared [78,79]. MRE is commercially available on 1.5T and 3T MRI scanners once suitable hardware is added in order to produce the requisite mechanical waves, and once specific software is installed for elastogram acquisition (Resoundant Inc., USA). Standardisation of 2D-MRE exists on three major vendors [80,81], although different shear wave frequencies are used outside the USA that are not FDA-cleared [82] and there is no consensus yet on the standards for ROI number, size or shape, which can add to measurement variability [83]. This could be overcome by dedicated freehand ROI selection under supervision of an experienced radiologist or by using additional software to aid in this process [84]. The accuracy of MRE for early fibrosis is reportedly superior to transient elastography (TE) but equivalent in cases of advanced fibrosis [34,85‐87], and MRE shear waves may propagate through small- and medium-sized ascites. MRE has a lower measurement failure rate than TE and has reportedly better repeatability [88]. However, iron deposition in the liver is a reported confounder for MRE that is significantly associated with measurement failure [77]. MRE is confounded by even mild iron overload, necessitating mpMRI with T2, T2* or, more recently, SE-EPI sequences [35,89,90]. Additional measurement of PDFF to evaluate steatosis has been attempted alongside MRE [36,40], as PDFF and T2* can be acquired within a single breath-hold. As with TE, MRE values are affected by chronic and acute inflammation, which can cause overlap in elastography values of patients with no or mild fibrosis [9,91]. Thus, high liver stiffness values can be obtained without any degree of fibrosis, resulting in low positive predictive value.
-
T1/corrected T1 (cT1): Modified Look-Locker inversion recovery (MOLLI) T1 maps provide diagnostic information in the heart, so that increased T1 can be diagnostic of oedema (increased tissue water) or increased interstitial space [92‐94], whilst increased extracellular volume is a powerful independent predictor of mortality in patients with severe aortic stenosis [95]. Similarly, the T1 of the water component is of diagnostic significance in the liver using MOLLI mapping or inversion recovery echo-planar imaging readouts to characterise tissue [28,68,96‐98]. Since iron is a ferromagnetic material, it can shorten tissue T1 and T2 relaxation times and this is further accentuated by the dependence of MOLLI T1 on T2 [99], with potential bias equivalent to one fibrosis stage when hepatic iron content increases from normal to high levels (1.0 to 2.5 mg/g) [100,101]. The confounding effect of iron on T1 mapping is corrected by a compensatory algorithm, based on the application of a multi-compartment model to simulate tissue and water environments in the liver during changes in iron content and in extracellular fluid (a proxy for fibrosis) [100]. For simplification the resulting cT1 is treated as a parametric component in this review, despite requiring T2* measurement. cT1 is commercially available as post-processing software (with T2* and PDFF, LiverMultiScan™, Perspectum, UK) and correlates with parenchymal fibrosis, inflammation and ballooning [16,28,31,32,76,102]. cT1 shows low measurement failure rates, high repeatability and reproducibility that are superior to those of elastography techniques in both published and preliminary data [42,80,88]. Fat has some additive effect on MOLLI T1 measurements at 3T [103] and by extension on cT1, but optimisation of MOLLI sequence parameters may be used to manage these biases, for example by use of asymmetric echo times during bSSFP [104]. Correlation of cT1 with histological disease features is maintained even after controlling for steatosis [76].
-
Diffusion-weighted imaging (DWI): Quantitative measures of diffusion can be produced by measuring the magnitude (apparent diffusion coefficient; ADC) and directionality (fractional anisotropy) of diffusion. The accumulation of steatosis, inflammation and fibrosis can lead to changes in water diffusion and these can be measured using various DWI techniques. Whilst mainly applied clinically in focal lesion characterisation, recent developments have potential utility in viral hepatitis and staging fibrosis [105‐110]. Limitations include lack of standardisation with inconsistencies reported for field strength [111] and B values [112,113]. Imaging homogeneity artefacts can be improved with simultaneous multi-slice respiratory-triggered acceleration (SMS-RT-DWI) [114]. Another DWI approach with promise is IVIM (intravoxel incoherent motion) that utilises diffusion imaging methods to explore both microcirculatory water motions and diffusion. Although specialist processing tools are required, recent studies indicate IVIM may further enhance fibrosis staging but this requires validation beyond focal disease [115‐118].
Technology characteristic | R2 (FerriScan®) | cT1-T2*-PDFF (LiverMultiScan™) | MRE |
---|---|---|---|
Turnaround time | 2-day service | 1 h service | Point of care |
Required hardware | 1.5T MRI scanner | MRI scanner, multiple field strengths and manufacturers | MRI scanner and driver device to generate mechanical waves |
Regulatory clearance | CE; TGA; FDA(510 k) and as Companion Diagnostic Device | FDA (510 k); CE; TGA; SMDR; NZ: listed on MEDSAFE | FDA (510 K) for 2-D MRE only |
Standardisation of hardware (CoV) | Manufacturer-based biases reported for R2 and R2* [188] | For cT1 3.3% (Siemens, Philips) [88] For T2* 6.6% (Siemens, Philips) [88] For PDFF 0.8% (Siemens, Philips) [88] | |
Diagnosis of iron overload | |||
Diagnosis of NASH and disease activity | Not applicable | cT1 AUROC for NAFLD: 0.93 [102] cT1 AUROC for ballooning: 0.84 [adapted from [32]] cT1 AUROC for NAS ≥ 5: 0.74 [102] | AUROC for NASH: 0.58 [85] |
Diagnosis of steatosis | Volumetric fat fraction of liver tissue (% fat) rather than PDFF available as HepaFatScan | PDFF AUROC for steatosis grades: ≥ G1: 0.93 [85] ≥ G2: 0.96 [85] ≥ G3: 0.94 [85] | |
Diagnosis of fibrosis | cT1 AUROC for fibrosis stages: ≥ F2: 0.63–0.79 ≥ F3: 0.62–0.74 | AUROC for fibrosis stages: ≥ F2: 0.83 2D-MRE [85] ≥ F3: 0.96 2D-MRE [85] F4: 0.91 [82] | |
Diagnosis of high risk | No | AUROC for NASH or fibrosis ≥ F2: 0.83 [102] | |
Coverage of liver | Whole slice analysis over 11 slices based on ROI | Whole slice analysis over 1–4 slices based on ROI or segmentation | Whole slice analysis over 4 slices based on ROI |
Measurement failure rate | Not reported | ||
Repeatability (CoV) | 15–21% [51] | For cT1 1.7–3.3% For T2*2.6- 5.5% For PDFF 0.8–8.8% | 11.0% [88] |
Confounded by iron | Not applicable | ||
Confounded by fat | Yes, for cT1 but can be managed by altering sequence parameters | Yes, in paediatric populations [161] | |
Confounded by comorbidities | Yes, fibrosis [54] | Yes, inflammation and fibrosis both increase cT1 [16] |