From the early 3rd trimester to around birth, the cerebral cortex undergoes dramatic microstructural changes including dendritic arborization that disrupts the radial scaffold, a well-organized columnar organization. Decrease of cortical fractional anisotropy (FA) derived from DTI has been well documented [1-3]. In this study, we hypothesized that non-Gaussian water diffusion properties (e.g. mean kurtosis or MK) from diffusion kurtosis imaging (DKI) [4-5] offer unique and complementary information on cortical microstructural changes during this period. The spatiotemporal changes and patterns of cortical FA and MK from 32 to 41 postmenstrual weeks (PMW) were revealed. Specifically, distinctive cortical MK and FA maps at each PMW during preterm development were characterized. Cortical MK and FA clustering patterns were revealed. The maturational curves of gyral level cortical MK and FA were also delineated to test the hypothesis.Subjects and data acquisition: 76 normal preterm and term neonates (47 Male and 29 Female; gestational ages of 32 to 42 PMW; 37.1±2.5 PMW) were recruited and scanned. Diffusion weighted images (DWIs) were acquired from a 3T Philips Achieva system with following imaging parameters: single-shot EPI sequence (SENSE factor = 2.5) without sedation, FOV=168/168/96mm, imaging matrix=112x112, axial slice thickness=1.6mm without gap, 30 directions; b-values=1000 and 1600 s/mm², repetitions=2, imaging time=18minutes. Of the 76 subjects, 27 (20 Male and 7 Female; 36.7±2.4 PMW) were scanned with two high b-values (1000 and 1600 s/mm²). Kurtosis and tensor fitting: The tensor fitting was conducted with DWI of b=1000s/mm² after motion and distortion correction to obtain the FA map. After DWIs of b=1600 s/mm² were corrected for motion and distortion, kurtosis was fitted using in-house software in MATLAB to obtain the MK map. Extraction of the cortical skeleton and measuring cortical FA and MK at the core of cerebral cortex: With TBSS of FSL (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/TBSS), cortical skeletons were extracted. Cortical tissue probability map was obtained based on the contrast of the MD map. At a given cortical skeleton voxel, the offset between the cortical skeleton and the core of the cerebral cortex was corrected by projecting the FA (or MK) from surrounding voxels with the highest gray matter probability to the cortical skeletons [3]. The FA and MK at the core of the cerebral cortex were averaged for all subjects at the same PMW and mapped to a template cortical surface. Measuring gyral level FA and MK at the cortical skeleton: The 56 gray matter regions of the JHU neonate atlas [6] were transformed to the cortical skeleton of each subject to measure the gyral level cortical FA and MK. Clustering of cortical MK and FA: Non-negative matrix factorization (NNMF) was conducted on the MK and FA measured at the cortical skeleton with the number of clusters set to 4 [7]. As shown in Fig. 1 and Fig. 2, decreases of both cortical MK and FA from 32 to 41 PMW can be observed during the 3rd trimester. However, the averaged cortical MK map is distinctive from the averaged FA map at the same age. For example, highest MK (Fig. 1a) is located at occipital region while highest FA (Fig. 2a) is located at prefrontal region at 33 PMW. The MK cluster distribution (Fig. 1b-1c) is also different from FA cluster distribution (Fig. 2b-2c). Compared to FA clusters, more separation in MK clusters can be appreciated. In Fig. 3, FA is best modeled with a biphasic piecewise linear fitting with a significant FA decrease from 32 to 37 PMW and relatively flat FA from 37 to 42 PMW. Significant age-dependent MK decreases were found at the precentral and postcentral gyrus, while there was no significant change in FA for these cortical areas. The cortical MK and FA maps are distinctive at the same age, as well as MK and FA maturational curves for the same cortical region. In addition, decrease patterns of cortical MK are different from those of cortical FA with MK clusters more clearly separated. These findings indicated that unique and complementary information on cortical microstructural changes can be offered by cortical MK measurements during this period. Cortical FA is sensitive to cellular processes such as dendritic arborization and disruption of radial glial scaffold [e.g. 1-3] which take place early in primary motor (precentral gyrus in Fig. 3a) and somatosensory (postcentral gyrus in Fig. 3b) regions, resulting in lower FA values in these regions as early as around 32PMW. Significant MK decreases all over the brain may be related to continuous decrease of diffusion barriers possibly associated with continuous decrease of neuronal density from 32 to 42PMW [8].