PurposeDiffusional kurtosis imaging (DKI) measures the deviation of the displacement probability from a normal distribution, complementing the data commonly acquired by diffusion MRI. It is important to elucidate the sources of kurtosis contrast, particularly in biological tissues where microscopic kurtosis (intrinsic kurtosis) and diffusional heterogeneity may co‐exist.MethodsWe have developed a technique for microscopic kurtosis MRI, dubbed microscopic diffusional kurtosis imaging (µDKI), using a symmetrized double diffusion encoding (s‐DDE) EPI sequence. We compared this newly developed µDKI to conventional DKI methods in both a triple compartment phantom and in vivo.ResultsOur results showed that whereas conventional DKI and µDKI provided similar measurements in a compartment of monosphere beads, kurtosis measured by µDKI was significantly less than that measured by conventional DKI in a compartment of mixed Gaussian pools. For in vivo brain imaging, µDKI showed small yet significantly lower kurtosis measurement in regions of the cortex, CSF, and internal capsule compared to the conventional DKI approach.ConclusionsOur study showed that µDKI is less susceptible than conventional DKI to sub‐voxel diffusional heterogeneity. Our study also provided important preliminary demonstration of our technique in vivo, warranting future studies to investigate its diagnostic use in examining neurological disorders.