Do Early MRI Signals Predict Lesion Size in a Neonatal Stroke Rat Model?

SUMMARY: In this study, we compared lesion size by using VADC and VT2 at 0, 2, 5, 24, and 48 hours and histologic lesions at 48 hours in a P7 rat stroke model. The best correlation between VHISTO and VADC was at H0, and between VHISTO and VT2, at H2-H5. Early MR imaging signals allowed excluding “no-lesion” and “no-reflow” animals to help standardize this neonatal stroke model and predict lesion size.

N euroimaging, particularly MR imaging, could be a tool for a rapid comprehensive assessment in acute stroke with the potential to guide treatment decisions for a better clinical outcome. One of the key issues in stroke assessment is to choose the best end points that appropriately indicate outcome. In experimental animal research, final lesion size has been used consistently as an outcome. Infarct sizes assessed by T2WI have been shown to correlate well by some and to underestimate histologic damage by others. 1 In the developing brain, authors have reported that MR imaging in P7 rat pups submitted to stroke 2 or hypoxia-ischemia 3,4 seems correlated to final histologic lesions measured 24 hours or 3-10 days after the injury, without early MR imaging time-point evaluation.
The aim of the present study was to develop a preclinical design including routine MR imaging coupling ADC maps, T2 maps, and MRA. We suggested that this strategy could improve the statistical power of trials using our model by diminishing the variability of the outcome assessment. Early MR imaging could 1) confirm initial ischemia, 2) confirm arterial reflow, and 3) accurately predict the final histologic outcome at 48 hours.

Perinatal Ischemia
Care and use of experimental animals followed French and European Community guidelines. Ischemia was created in 24 Wistar P7 rats (Janvier, Le Genest Saint Isle, France) of both sexes. 5 Briefly, anesthetized (chloral hydrate, 350 mg/kg, intraperitoneally) rats were exposed to left MCA electrocoagulation followed by 50-minute transient left common carotid artery occlusion. During ischemia and recovery the pups were placed in a humidified incubator at 37°C.

Outcome Assessments
The volume of the lesion was assessed by using histology at H48, ADC maps, and T2 maps from H0 to H48. V HISTO was delineated by a classic visual method on cresyl violet-stained coronal sections (50-m thick, 500-m intervals). V ADC and V T2 were defined by applying a threshold (mean Ϫ 2 SDs) of the whole contralateral hemisphere ADC and T2, respectively. Volume measurements were expressed as a percentage of the ipsilateral hemisphere.

Statistical Analysis
All results are given as mean Ϯ SD. The correlation and regression were performed with a standard statistical software package (SSYSTAT 3.1, Systat Software, San Jose, California) by using a Spearman rank-order correlation and a linear regression, respectively.

RESULTS
Four animals of 24 (17%) were excluded as no-lesion. All exhibited MCA interruption on H0 angiography, without a significant DWI lesion. Two animals of 24 (8%) were excluded as "no-reflow" as determined on MRA. Eighteen animals were then analyzed (Fig 1), and V ADC , V T2 , and V HISTO were calculated (Fig 2). An area in the ipsilateral frontoparietal cortex with low V ADC was maximal at H0 in almost all the animals (Fig 1). V ADC at H0 was different from 6.6%-32.1% of the ipsilateral hemisphere. A significant correlation between V ADC at H0 and lesion volume (V HISTO ) at 48 hours was obtained (slope: 1.09, r ϭ 0.92, P Ͻ .001; Fig 2). In some animals, V ADC slightly decreased between H2 and H5, whereas it remained similar in the others. At 24 and more markedly at 48 hours, all V ADC values were highly reduced (0.2%-8.3% ; Fig 1). Even if hints in the deeper cortical and subcortical layers (Fig 1) were visible on T2 maps as soon as H0 (in few animals), the T2 lesion was clearly identifiable in all animals at H2 (4.4%-26.7%). A significant correlation between V T2 and V HISTO was obtained at H2 (slope: 0.97, r ϭ 0.91, P Ͻ .001) and H5 (slope: 1.04, r ϭ 0.91, P Ͻ .001; Fig 2). This correlation remained at H24 but with a higher slope (1.15, r ϭ 0.89). Then V T2 slightly decreased at H48 (r ϭ 0.888). The range of V HISTO was from 3.5% to 24% with a mean of 13.5 Ϯ 5.7%.

DISCUSSION
This study showed that routine MR imaging could be a toolbox able to standardize neonatal stroke models, first by excluding no-lesion and no-reflow animals in the first MR angiography, and second by predicting histologic outcome early in a neonatal stroke model.
The first set of MR images at H0 was used to exclude no-lesion animals. These animals probably developed efficient cortical arterial anastomoses during ischemia as we previously demonstrated by using sonography. 6 MR imaging also allowed excluding no-reflow animals because the absence of recanalization was shown to block the development of further edema. 7 All animals exhibited a quite reproducible time course of ADC or T2 volume. As previously reported, V ADC reached a maximum just after reperfusion. 7 The best-correlated V ADC time point with V HISTO was H0. This impressive correlation lets one think that as soon as 15 minutes after reperfusion, "the Note that low ADC was observed in the lesioned frontoparietal cortex compared with the contralateral side. Hints of T2 were observed as early as H0 in the deeper cortical layers. At H48, marked T2 signals are still detected in the deeper cortex. die was cast," and almost the whole interindividual variability of animals was already determined. In contrast, at H2 and H5, the slope of regression was superior to 1, suggesting that V ADC overestimated histology volume. Conversely, at 24 and 48 hours, correlation between ADC and histology became less strong and the slope coefficient decreased, leading to an underestimation of the final lesion.
As previously reported, T2 lesions appeared as soon as 2 hours after ischemia. 7 Most interesting, we observed that the correlation of V T2 with V HISTO was strong as early as H2 and H5, with a slope of regression close to 1, suggesting good overlap between T2 signals and histologic lesions. Thereafter, V T2 overestimated lesions at both 24 and 48 hours.

CONCLUSIONS
This whole strategy was not reported in the literature because only parts of studies reported exclusion of no-lesion animals with intraischemic MR imaging. 8 In addition, early MR imaging signals appear to be good surrogates for final stroke lesions in the developing brain.