We present Spatial PAttern Recognition via Kernels (SPARK) as an effective statistical tool to identify genes with spatial expression patterns in spatially resolved transcriptomic studies. A particular feature of SPARK is its ability to produce calibrated p-values for spatial analysis.