To extract 3D brain tumors and tissues, the paper develops a segmentation approach based on U-Net neural network integrating with surface area and volume measurement. From the research-oriented open source we take the simulated brain database and BraTS database as input data. To obtain the neuron's weighting in the neural network, we use this training data to train the U-Net neural network. The obtained U-Net neural network is to be used in our extraction method, and the extraction of the brain tumors and tissues will be executed continuously from a series of brain images.