The invention discloses an adaptive layered neighborhood radius method for classifying abnormal brain tissue medical records, which comprises the following steps: first constructing a four-layer structural segmentation framework based on the adaptive neighborhood radius, and assigning each layer to a neighborhood The effective solution of the radius is to adaptively adjust the neighborhood radius through iterative calculations; then, the interaction of the neighborhood radius within the same layer and the cascade operation of the neighborhood radius between different layers are performed on the medical records of abnormal brain tissue. The present invention can effectively extract the attribute characteristics of different regions inside the medical records of abnormal brain tissue, improve the classification efficiency of different longitudinal cortical surface labels of brain tissue, and have good effects on early prevention of abnormal brain diseases and delaying the onset of brain diseases.