Editorial for “Prospective Comparison of <scp>DWI</scp> ‐Derived Virtual <scp>MR</scp> Elastography and Conventional <scp>MR</scp> Elastography in Metabolic Dysfunction−Associated Steatotic Liver Disease and Healthy Volunteers

The determination of liver fibrosis severity by non-invasive tests represents an important objective in patients with liver diseases related to metabolic dysfunction [1]. Toward this objective, the measurement of mechanical properties with MR elastography (MRE) constitutes an essential tool [2]. Indeed, deposition of collagen fibrils is known to cause tissue stiffening, which can be sensitively picked up with MRE. However, measuring mechanical properties in tissues entails evaluating the stress–strain relationship, which imposes mechanically stressing the tissue throughout the depth of the targeted organ [3]. The resulting deformation is recorded with phase-encoded MR images, and is used to compute maps of mechanical properties. In MRE, stressing deep-seated organs can be achieved with mechanical waves propagated into the patient by an external mechanical transducer synchronized to the acquisition sequence. However, the use of an external transducer entails additional complexity, cost, and mild discomfort to the patient. Therefore, alternate methods for assessing liver fibrosis severity are desirable. Recently, Le Bihan and colleagues proposed a method exploiting anomalous diffusion and the kurtosis model for liver fibrosis quantification [4]. The method relies on a non-Gaussian effect of collagen fibers on water diffusivity to stage hepatic fibrosis in patients. Only two b-value images were acquired, followed by the calculation of an “sADC” index (for “shifted apparent diffusion coefficient”). Subsequently, an empirical calibration of sADC values against MRE-measured stiffness values was performed to enable the estimation of liver stiffness directly from diffusion-weighted imaging data. This method was initially established and further validated in a larger cohort of patients with liver fibrosis of predominantly viral origin, in two studies [5]. Subsequently, Hanniman and colleagues used the diffusion kurtosis sADC method in a population of patients with metabolic dysfunction-associated steatotic liver disease (MASLD) [6]. Because of high hepatocytic fat levels, Hanniman and colleagues proposed a fat-corrected adaptation of the approach from Le Bihan and colleagues and used liver biopsy rather than MRE as the reference examination. However, diffusion kurtosis−based fibrosis estimates were not correlated with fibrosis stage in these patients. In a response letter to the editor, Le Bihan and colleagues hypothesized that this lack of correlation may have been explained by Hanniman and colleagues not having utilized a b = 0 image in their calculation. Le Bihan and colleagues also proposed an alternate model for sADC computation in the presence of fat [7]. In their paper ‘Prospective comparison of DWI-derived virtual MR elastography and conventional MR elastography in metabolic dysfunction-associated steatotic liver disease and healthy volunteers’, Volniansky and colleagues have used the fat-corrected version of the sADC computation in a cohort of 43 MASLD patients and 11 volunteers (dummy quote) [8]. Besides evaluating the reproducibility of the approach, the authors also evaluated the diagnostic performance for staging liver fibrosis with liver biopsy as the reference method. Their results indicate that despite fat correction, no correlation could be found between MRE-derived stiffness and the diffusion-based markers. Moreover, only MRE had significant area under the receiver operating curve for hepatic fibrosis staging. These results are in concordance with those obtained in our group, where no significant relation could be observed in MASLD patients between fat-corrected diffusion-based metrics and the liver fibrosis stage determined with liver biopsy as reference [9]. This absence of link between sADC and fibrosis in MASLD patients could have several causes. The presence of intracellular lipid vesicles is likely to yield complex behavior in the kurtosis regime of diffusion. The microscopic origin of diffusion kurtosis remains an active field of study, and the kurtosis equation is currently viewed as a perturbation derivative of the diffusion equation where diffusion kurtosis is a phenomenological parameter [10]. One possibility is that the currently proposed fat correction models, or the assumptions on which they rely, could be insufficient to accurately depict such complex behavior. For instance, the model implicitly assumes that the signal originates additively from two compartments without cross-terms. The fat compartment is also assumed to have no or negligible diffusion. The relaxation times for water and fat and the fat suppression efficiency factor of the sequence, are also assumed to be spatially invariant and constant in all patients. While these assumptions are reasonable, in MASLD they may still prove insufficient to accurately distinguish between the effects of steatosis and fibrosis on the diffusion of water molecules and ultimately on the signal. This may obscure the relationship between fibrosis and sADC that is so prominent in viral hepatitis patients. Furthermore, tissue structure and necro-inflammation differ between viral hepatitis and MASLD. Finally, the different reference methods (MRE in the studies on viral hepatitis and histopathology in the studies on MASLD), could also have contributed to the differing outcomes. Overall, the work of Volniansky and colleagues further establishes that the potential role of the sADC technique in assessing liver fibrosis in primarily viral hepatitis patients could not be directly transferred to patients with MASLD. This work illustrates the difficulties that are encountered in finding a universal imaging marker of liver fibrosis. This work was supported by Agence Nationale de la Recherche (ANR-17-RHUS-0009, ANR-20-CE19-0005) and doctoral grant from Université Paris Cité (doctoral school 563 MTCI).