This work is of useful for scientists and radiologists interested in diffusion MRI, in particular, in methodology of group comparisons. Diffusion tensor (DTI) and diffusion kurtosis imaging (DKI) are powerful tools extensively used to investigate brain microstructure and connectivity. The DTI/DKI metrics are often used for between-group comparisons and studying the correlations with age or severity of pathology. However, the methodology of group comparisons is not sufficiently well established. Popular methods include region-of-interest (ROI) analysis [1], voxel-based whole brain analysis [2], and the track-based spatial statistics (TBSS) [3]. Each method suffers from specific problems reducing its statistical power or sensitivity. In particular, difficulties of the ROI-based analysis are associated with the necessity of manual (therefore, user biased) time-consuming delineation of the ROIs and large inter-subject variability. To overcome these difficulties, the atlas-based segmentation is used to automatically define the ROIs by co-registration of the subject- and atlas-spaces. However, misregistration and various distortions lead to significant partial volume effects (PVE) [4,5]. Therefore, automatic atlas-based segmentation is prone to errors diminishing its accuracy. An alternative approach, TBSS, attempts to overcome misregistration problems by skeletonizing WM and considering only the voxels of the mean fractional anisotropy skeleton. However, since the skeleton is formed only by a small fraction of voxels, potentially useful information occurs significantly reduced. The purpose of this work was to develop a novel method that would reduce PVE in the atlas-based analysis, on the one hand, but retain more voxels for between-group comparisons, on the other. PVE are especially crucial when white matter (WM) ROIs appear contaminated by voxels belonging to gray matter (GM) or cerebrospinal fluid (CSF). Here, we propose a new diffusion-Kurtosis-Informed Template (KIT) that allows one to reduce contamination of WM ROIs by PVE from neighbouring GM and CSF regions by considering specific features of the diffusional kurtosis histograms. DKI measurements were performed with a whole-body 3T Siemens MAGNETOM scanner for 2 groups of healthy subjects of different age. Typical DTI/DKI metrics (mean (MD), axial, and radial diffusivity, fractional anisotropy, mean (MK), axial, and radial kurtosis, kurtosis anisotropy were determined on the voxel-by-voxel basis in the whole brain and averaged over 20 WM anatomic regions provided by JHU WM tractography atlas available in FSL.KIT for atlas-based comparison. Fig.1a shows the histogram of a typical MK map (Fig.2). The data points are fitted by a sum of the 2 Gaussian distribution functions (see Eqs. 13,14 and Fig. 10 in Ref. [5]). Such a double-peak histogram is characteristic for brain regions containing voxels in both WM and non-WM (GM and/or CSF) regions: the component with smaller MK values is attributed to non-WM, the component with higher values to WM [5]. The Gaussian shape of the MK distribution in WM (red curve, Fig. 1b) is favourable for an easy and formalized determination of the MK value (MK_HHL) at its half height (HH). We determine the KIT by selecting the voxels for which the following conditions hold: a) FA > 0.2 and b) MK > MK_HHL. The KIT is then applied for all evaluated metrics. The condition (a) is widely used to roughly separate WM and non-WM. However, as the scatter plots in Figs. 1c and 1d clearly show, this condition is insufficient to eliminate a significant amount of voxels exhibiting low MK (see quadrant IV in Fig. 1c) and/or high MD (see blue data points in Fig. 1d) paralleled by relatively high FA (which can partially be encountered due to evaluation errors). These voxels very likely belong to non-WM and should be excluded from comparisons. Voxels that finally form the KIT are shown by data points in quadrant II in Fig. 1c and by data points in red in Fig. 1d. The KIT is additionally visualised in Fig. 3 in red, whereas the voxels eliminated by the constraint (b) are represented in green. One can clearly see that the majority of the eliminated voxels are really located at the interfaces of WM and non-WM regions justifying our simple approach as robust and effective. We also demonstrated the efficiency of the KIT by its application to an atlas-based comparison of 2 differently aged groups of healthy subjects. We showed that both the magnitude of the between-group differences of the DTI/DKI metrics and the corresponding significance levels become greatly larger after an application of the proposed method. We demonstrated a simple, robust, and efficient method to reduce PVE in the atlas-based analysis, in general, and in between-group comparisons, in particular, by using the novel diffusion-kurtosis informed template.