Parametric response mapping as a new MRI approach for the characterization of post-traumatic cerebral edema

Purpose Cerebral edema is usually assessed with mean ADC value obtained with MRI. We aimed to use parametric response mapping (PRM), a voxel-based analysis, in a rat model of traumatic brain injury (TBI) to characterize the cerebral edema and to compare findings to those obtained with the classical global mean ADC analysis. Ultimately we wished to test whether early PRM analysis could identify the vasogenic or cellular edematous pattern two hours after trauma. Materials and Methods Experiments were conducted in a rat model of TBI in accordance with French Government guidelines. Eighteen traumatic brain-injured rats (TBI group) were compared to 7 sham-operated rats (Sham group). Diffusion-weighted images were acquired before and immediately (H0), 60min (H1) and 120min (H2) after the trauma. Two regions of interest (ROI) including cortex (cortical-ROI) and the whole brain (brain-ROI) were manually delineated. Results Averaged ADC value in cortical-ROI was significantly reduced at H2 in the TBI group versus Sham group (715±24 vs 781±16 μm2/sec; p=0.04), not in brain-ROI (748±27 vs 801±18 μm2/sec; p=0.32). By contrast, PRM values were significantly changed in these two ROIs at as soon as H1 in the TBI group (27.5±4% of the voxels beyond the PRM threshold in TBI group vs 5.1±1% in sham group; p<0.01). PRM was able to identify the nature of post traumatic cerebral edema: vasogenic, cellular or mixed pattern. ADC data obtained immediately after TBI and analyzed using PRM correlated with ADC data obtained at H2. Conclusion The PRM approach is more sensitive than averaged ADC at detecting and identifying post-traumatic brain edema. It is able to PRM might be a promising tool to individualize the management of patients with posttraumatic brain edema.

controlled cortical impact models obtained prominent cellular edema within the core lesion 81 and vasogenic edema in the surrounding area (9,10)).   The PRM analysis was conducted as previously described by Galbàn et al. (

178
The two groups of rats were comparable in terms of their physiological data at H.ref, 179 except for PaO 2 which remained physiologic or supra physiologic in both group (Table 1).  Fig. 1a). This difference was not found when the 189 brain-ROI was considered: 748±27 µm²/sec (TBI group) vs. 801±18 µm²/sec (Sham group)

190
(p=0.32) (Fig. 1b). Except for the cortical-ROI at H2, mean ADC values were comparable 191 between the TBI and the Sham group for all ROI and at all time-points. an initial phase of vasogenic edema followed by a phase of cellular edema (Fig.2 (Fig. 3). The total number of voxels found beyond the 207 PRM threshold (i.e. red + blue voxels) were 27.5±4% vs 5.1±1%, respectively (p<0.001).

250
Conversely, the cellular edema subgroup presented a sustained proportion of more than 25% 251 of blue voxels with less than 8% of red voxels (Fig. 5) Altogether, after TBI, rats exhibit a mixed pattern of red and blue voxels, with a clear 258 predominance of one or the other. Averaging ADC values in the whole brain failed to detect the occurrence of brain 267 edema in our rat model of diffuse TBI. A voxel-wise basis approach (PRM) could however 268 detect the nature of post-traumatic brain edema, i.e., vasogenic, cellular or a mixed pattern, 269 and is able to provide temporal and spatial information concerning evolution of this edema. Moreover, the mean-value approach is limited to one quantitative estimate (the mean ADC in 290 this case) whereas PRM also allows the measurement of the proportion of voxels which vary 291 above a predefined ADC threshold. 292 We were surprised to find almost no difference between the TBI and the Sham- restricted to specific brain areas usually chosen for ROI, such as the cortex or striatum, but 302 could be spread throughout the whole brain (Fig.2). The spatial and temporal evolution of 303 edema could thus be estimate by PRM within every animal. The focus on specific ROI could 304 therefore lead to misinterpretations, eluding the complexity of post-traumatic cerebral edema.

305
One important finding of our study is the presence of red voxels in almost every rat.

306
This elevation of ADC could be transient (Fig.2. Nevertheless, a PRM analysis of early ADC data was useful in estimating the edematous 319 pattern two hours after diffuse traumatic brain injury. This finding could be relevant with