This paper introduces the data processing of functional magnetic resonance imaging (fMRI) of brain based on statistical parametric mapping (SPM). Because of its excellent spatial resolution and temporal resolution as well as its noninvasive approach with no involvement in the injection of radioactive materials, the fMRI of brain has been widely used by brain researchers in the world. There are four steps in the processing, i. e. reading the raw data, pre-processing, statistical analysis of the data and model estimation. The pre-processing includes motion correction, normalization, spatial smoothing and temporal smoothing. The statistical analysis of the data includes building the general linear model. Model estimation includes serial T-test and inspecting the activation region. In this paper, emphases are laid on the new methods in every processing.