1168 Water Diffusional Kurtosis Imaging (DKI) Analysis of Ischemic Stroke Model in Young Rats Renaud NICOLAS 1 , Gérard RAFFARD 2 , Stéphane SANCHEZ 2 , Florent AUBRY 1 , Eric PETERSON 2 , Isabelle BERRY 1 , Pierre CELSIS 1 , and Bassem HIBA 2 1 Imagerie cérébrale et handicaps neurologiques; UMR 825, INSERM, F-31059 Toulouse, France, 2 Centre de Résonance Magnétique des Systèmes Biologiques,UMR 5536, CNRS, F-33076 Bordeaux, France INTRODUCTION : Diffusional Kurtosis Imaging (DKI) is a clinically-accessible, NMR model of non-gaussian properties of water diffusion in tissue (considering b-factors up to 3000 s/mm²) [1]. Although based on a different hypothesis, the Statistical Model of Diffusion Imaging (SMDI) [2,3] is equivalent to the DKI model [1,2,3]. Although ischemic stroke has been studied extensively with gaussian diffusion models, it has only recently been shown in humans that the apparent kurtosis excess (K app ), that depend on the diffusion coefficients dispersion, increases in ischemic regions [5]. The excess Kurtosis of an unique water pool in a heterogeneous media is expected to be sensitive to both water restriction effects and exchange when varying diffusion time t d [1, 3]. MATERIALS AND METHODS : Stroke model : 7 juvenile rats (7 days, ~20 g) were submitted to Middle Cerebral Artery Occlusion (MCAO) followed by a 2 h incubation in 8 % O 2 , 92 % N 2 at 31-33°C. NMR acquisition : DWI-EPI images (357 μm x 312 μm, 6 averages, 10 slices of 1 mm thick, TE/TR=68/3200 ms) with diffusion gradients applied in R,S,P directions were acquired on a 4.7 T Brucker small animal scanner. A constant diffusion time (CT) Stejskal-Tanner MRI sequence was used with b-factors of 0, 500, 1000, 1500, 2000, 2500 s/mm² with a with diffusion gradient duration δ=4ms. Three different experiments were performed using gradient separation Δ=10,30,50 ms. Post-Processing : After registration and eddy-current correction, the direction independent S/S 0 pixel-by-pixel signal decay was interpolated to b-factors of 100, 200, 500, 750, 1000, 1500, 2000, and 2500 s/mm². After that, the S/S 0 images were non-linearly fitted by the medium b-values range simplified form of SMDI : S/S 0 =exp(-b.D app +½b².σ²) [4,5]. K app was computed as K app =3(σ²/D app ²) [3,4,5]. Statistical Analysis : Cortical ROIs of healthy and ischemic areas, traced on D app maps, were used as masks allowing the determination of D app and K app pixel-averaged spatial mean values, variances and histogram profiles. A non-parametric Kruskal-Wallis (rank sum) test, adapted for few-subjects study, was used to assess the statistical significance of the differences between ischemic and healthy tissue as well as the effects of diffusion time variations on pixel- averaged spatial mean, designed as <D app > and < K app >. RESULTS : For all diffusion gradient separation time (Δ=10,30,50 ms), the ischemic cortex is characterized by a reduction of <D app > (P=0.001745), and a rise of <K app > (P= 0.001745), when compared to healthy contralateral tissue, as illustrated by Fig. 1 and 2. Table 1 show the <D app > and < K app > values averaged for all rats studied. A reduction of 40 % for <D app > and a 55 % rise for <K app > were observed in the ischemic region compared to the healthy contralateral area. As shown in Fig 2., ischemic K app values are more dispersed than normoxic K app , ischemic D app or normoxic D app values. If D app histograms in healthy versus diseased tissue were well separated (Fig 2, top), there is an overflow of the measured K app values (Fig 2, bottom). No statistically significant differences were found (in both healthy and ischemic cortex) among <D app > and < K app > values obtained with the three different diffusion times. Table 1. Pixel-averaged intensity <D app > and <K app > values (given as mean ± total standard deviation) averaged over the 7 rats. Standard deviation include intra- and inter-subject dispersion around pixel mean value and subject mean value. Fig. 1 (left). Parametric images of 2 non-contiguous slices of one rat 3h following MCAO. D app (top) is displayed in μm²/ms, K app (bottom) is dimensionless. Fig. 2 (right). Slices averaged histograms of D app (top) and K app (bottom) corresponding to Fig. 1 images, for healthy cortex (black) and ischemic lesion (red). For Fig. 1 and 2, images and histograms from left to right correspond respectively to Δ=10,30,50 ms. DISCUSSION : For medium b-values, DKI and SMDI are equivalents models describing the non-gaussian diffusion of an unique water pool. Extending the fit to a complete SMDI model would possibly lead to the usability of a DKI-type model for b-values higher than 3000 s/mm² [2]. High K app are representative of strong dispersion of mean diffusivities, reflecting the structural complexity of tissue [3]. As observed in humans [5], K app is as sensitive as D app (or equivalently, clinical ADC) in delineating the extent of ischemic region in the cortex. K app also shows a spatial heterogeneity in the lesion compared to D app . In this preliminary study performed in a limited number of subjects as well as with a limited diffusion times range, spatial averages <D app > and <K app > for healthy and ischemic parenchyma do not present any significant changes when varying diffusion times. Histological assessment of tissues properties in the stroke lesion is in progress and would confirm if K app maps are able to differentiate internal tissue characteristics in the stroke lesion. REFERENCES : [1] Jensen J.H. and Helpern J. A, MRI quantification of non-Gaussian water diffusion by kurtosis analysis, NMR Biomed. 23:698–710 (2010) [2] Yablonskiy D. A. and Sukstanskii A.L, Theoretical models of the diffusion weighted MR signal, NMR Biomed. 23:661–681 (2010) [3] Novikov D.S, Kiselev V.G, Effective medium theory of a diffusion-weighted signal, NMR Biomed. 23(7):682-97. (2010) [4] Yablonskiy D.A, Bretthorst G.L and Ackerman J.H, Statistical model for Diffusion Attenuated MR Signal, Magn. Res. In Medecine 50:664-669 (2003) [5] Jensen J.H, Falangola M.F, Hu C, Tabesha A, Rapalinod O, Loa C and Helpern J.A, Preliminary observations of increased diffusional kurtosis in human brain following recent cerebral infarction, NMR Biomed. 24:452–457 (2011) View publication stats View publication stats