The traditional minimum spanning tree feature extraction method uses local quantifiable indexes to classify brain diseases, neglects the important role of low weight connections and clusters in information processing in brain network, resulting in the loss of some useful information in the network. Compared with other network features, these indexes show significantly lower feature validity and classification accuracy. In order to solve these problems, we took the minimum spanning tree topological index as the feature, to extract the feature on the local difference minimum spanning tree brain network, construct the classifier, and verify it on the depression patient data set. The experimental results show that this method can provide more effective features and improve the classification accuracy with respect to the traditional minimum spanning tree method. Further analysis shows that the new method proposed in this paper can provide an important reference for the construction of brain network and feature extraction, and also contribute to the research of medical assistant diagnosis and brain diseases.