The ADC calculated on most vendor MR platforms, which are in routine clinical use, assumes a mono-exponential relationship between the measured signal intensity and the diffusion-weighting (b value). However, in tissues, the signal attenuation with increasing b value is non-linear, with an increased signal attenuation at lower b value, which is ascribed to tissue microcapillary perfusion. By performing DW-MRI using multiple b values (including low b values of typically <200 s/mm2), it is possible characterize or estimate both the “true diffusion” from the “pseudodiffusion” (related to microcapillary diffusion) by applying the principles of IntraVoxel incoherent motion (IVIM [51]), which assumes a bi-exponential relationship between the measured signal intensity and the b value. Parameters derived using this approach include the perfusion fraction (f, a simplified view is that it represents the fraction of vascular flow in tissue), the pseudodiffusion coefficient (D*, the rate of vascular flow) and the diffusion coefficient (D, representing tissue water diffusivity), the product, fD*, providing an estimate of perfusion.
The ability to provide both diffusion and perfusion quantification using a single imaging study, without the need for intravenous contrast injection, appears highly attractive. Others see bi-exponential data fitting by IVIM analysis as a way of obtaining a potentially more robust and accurate measurement of tissue diffusion by accounting for the perfusion component. For these reasons, there has been a significant output of research in this area in the past few years, exploring DW-MRI derived diffusion and perfusion parameters for disease assessment [52–64].
Studies have already demonstrated that the renal cortex and medulla have different ADC values, and the f of the medulla has been proven to be lower than the cortex [65]. Recently, it has been shown that using IVIM analysis, the f together with the D showed the highest diagnostic accuracy for the diagnosis of clear cell renal carcinoma (Az = 0.78), and this was able to reliably distinguish between papillary cell carcinoma from cystic RCC [66]. In another study, the D showed a higher diagnostic accuracy than mono-exponential ADC in discriminating between clear cell and non-clear cell renal cancers [57]. IVIM analysis has been used in differentiating pancreatic carcinomas from the mimicking focal pancreatitis and surrounding normal pancreatic tissues [55, 67, 68]. More recently, the D was found to have a higher diagnostic performance compared with conventional ADC in discriminating between malignant and benign focal hepatic lesions (Az 0.96 versus 0.93) [53]. Furthermore, the f and D* were also significantly higher in hypervascular liver lesion compared with non-hypervascular lesions [53].
Despite these varied and positive results, a word of caution should be made. The IVIM analysis requires imaging acquired using multiple b values (typically six or more), with good image signal-to-noise ratio, to provide confidence in the results. Even then, because the diffusion model is fitted with three parameters, the pseudo-diffusion coefficient derived can be unstable, and some authors have thus chosen to fix this value a priori [68]. A study evaluating liver parenchyma and liver metastases has shown that the measurement reproducibility of the D is good in both normal liver and metastases, but in hypovascular metastases, the estimation of the f and D* is associated with very large measurement uncertainty (50 % or greater) [56•]. This poor measurement reproducibility in lesions with low f suggests that the technique may not be reliable in disease with inherently low pseudodiffusion phenomenon at low b values. For these reasons, IVIM analysis is still regarded as a research tool and has not filtered into mainstream clinical application.
More recently, other non-mono-exponential diffusion models have been applied models for DW-MRI evaluation, including Gaussian model [69], stretched exponential model [70] and the kurtosis model [71]. The IVIM model remains the most widely used, as it links the bi-exponential signal attenuation behavior in tissues to the specific underlying biological underpinning of microcapillary perfusion.
Interest in the kurtosis model (also called kurtosis diffusion in the literature) is, however, increasing. The diffusion kurtosis (K) measures deviation from Gaussian water diffusion behaviour, which is observed in free water or homogenous solutions. In tissues, the presence of microstructure and microcapillary perfusion results in deviation from this behavior. Hence, diffusion kurtosis (K) measurements can be thought of as measuring diffusion heterogeneity, thus reflecting tissue “complexity”, with higher K values having been reported in tumors [72, 73]. Early work suggests that the technique may be able to distinguish between tumor and benign pathologies; between low and high grade tumors; as well as between native and treated tumor tissues [72, 73]. Clearly, the role of diffusion kurtosis imaging (DKI) needs to be further clarified with future research, including studies to establish its measurement reproducibility across different body sites and tumor types.
No comments:
Post a Comment