1.
Liver ADC at 3.0T
Liver ADC values in healthy subjects have been reported either significantly increased [15] or decreased (but not significantly) [16] at 3.0 T as compared to those obtained at 1.5T. “Noise floor” from lower signal-to-noise ratio (SNR) at 1.5T and decrease in liver T2 relaxation time at 3.0T have been proposed as potential explanations for these opposite findings,
respectively.
By considering both planes of measurement,
the ADCs we showed in controls (Tab.
4-5) are consistent with those reported by Dale et al.
(2.26 and 1.56 x 10-3 mm2/sec at 0-400 and 0-800 sec/mm2,
respectively) [15] rather than by Rosenkrantz et al.
[16] (1.49 ± 0.47 and 1.12 ± 0.36 x 10-3 mm2/sec at comparable b-values set).
Considering that we used a TE of 66 msec instead of 76 msec [15-16],
which partially compensate for decreased T2 at 3.0T,
our data may support the hypothesis formulated by the former Authors,
suggesting that increasing SNR at higher field strength impacts on the ADC value of the liver.
It is questionable whether the use of ultra-high field strength may have affected measurements on cirrhotic patients.
Like explained in our EPOS 2011 exhibit #3051,
isotropy was shown in cirrhotic patients,
regardless of the anatomic plane of measurement and the b-values set.
Accordingly,
ADC values (Tab.
4-5),
are consistent throughout the liver.
However,
it remains questionable whether 1.5 or 3.0T provide more precise estimation of liver ADC,
considering that many other confounding equipment-related,
patients-related and DWI protocol-related factors currently limit its reproducibility [4].
2.
Diagnosis of hepatic cirrhosis
The accumulation of extracellular matrix component characterizing fibrosis theoretically reduces water diffusion within the liver,
i.e.
parenchymal ADC [5].
This assumption has been verified by a number of studies [6-9] and in our series,
since ADC in cirrhotic patients was shown lower than in controls (Tab.
6).
Nonetheless,
by using higher b value set (b=0-800 sec/mm2) we showed that DWI is unreliable in differentiating between healthy and cirrhotic livers,
because the ADC lowering in cirrhotics resulted not statistically significant at one of two planes of measurements,
and along two of three gradient directions (y and z). Note that ADC at this b-value set was assumed to be approximated to the true diffusion coefficient D.
Our data match with previous observations on animal model and humans,
suggesting that the so called perfusion-related diffusion,
or fast component of diffusion (D*) is larger in controls than in patients with liver fibrosis [10-11].
By virtually eliminating the effects of D*,
i.e.
by calculating the ADC with b-values higher than 400-500 sec/mm2,
the difference between cirrhotic and healthy livers is lost after proper correction for image noise [5].
Thus,
perfusion rather than properly said diffusion is advocated to represent the differential factor.
Furthermore,
ADC decreased as the maximum b-value increased from 400 to 800 sec/mm2 (in most cases with a p<0.01),
but at a lesser degree in cirrhotic patients than in controls (Tab.
4-5).
Although the assessment of D* was not possible in our model,
we performed indirect estimation by calculating the volume fraction f (corresponding to the volume of water flowing into capillaries within the voxel) [11].
Differently from Luciani et al.
[11],
and more coherently with the decrease of D* these Authors showed,
we observed that f is reduced in cirrhotic patients as compared to controls,
regardless of the gradient direction and site of measurement (Tab.
7).
Inhomogeneity in liver fibrosis distribution may explain why differences were not always statistically significant.
Our findings are in accordance with well documented reduction of blood volume in cirrhosis [17].
3.
Final consideration
In our opinion,
our findings further emphasize the need for a still lacking,
reliable radio-pathological correlation between fibrotic or cirrhotic changes and parenchymal ADC.
Targeted studies are needed to this purpose,
possibly by using diffusion tensor imaging (DTI) [18].
At the state-of-the-art situation,
DWI is probably technically immature to provide reliable information on liver fibrosis and cirrhosis.
Limitations
First,
we estimated liver ADC by using two b-values sets,
with two b-values each (0-400 and 0-800 sec/mm2).
Acquisition with multiple b-values might have improved the accuracy of ADC measurements [18],
especially because – as discussed above – the influence of ultra-high field strength on ADC is still undetermined,
and led to perfusion D* estimation [11].
We did not use multiple b-values to avoid: a) excessive increase in acquisition time,
that was already longer than usual (nominally about seven minutes) because of the application of sequential gradients along three different directions; b) additional image degradation at longer acquisition time (see discussion above).
Second,
similarly to previous Authors [15-16],
we used a maximum b of 800 sec/mm2,
i.e.
lower than feasible at 3.0T.
This choice was based on the assumption that,
irrespective of the field strength,
influence of perfusion D* should be minimal in determining the ADC at the b-value set of 0-800 sec/mm2 [11].
Even if the approximation to the true diffusion coefficient D would have been inaccurate,
the general trend of our results is not affected.
Besides,
we tried to avoid image degradation inherent to larger b-values.