Medical image classification method based on supervision graph regularization and information fusion

The invention provides a medical image classification method based on supervision graph regularization and information fusion. The medical image classification method is used for distinguishing pMCI and sMCI. According to the method, firstly, feature selection is performed based on auxiliary data, and effective original features are reserved for subsequent operation; the method comprises the following steps: firstly, extracting features based on supervision graph regularization and information transfer (SGRIP), and in the process, adaptively optimizing a neighbor relation of samples in a projection space by utilizing category label information of auxiliary data, so that the extracted features are more discriminative. And thirdly, the extracted SGRIP feature is fused with a mini-MentalStateExamination (MMSE) score, and an apolipoprotein E4 allele (apolipoprotein E4, APOE4), so that the diversity of the feature is enriched, and the method is suitable for the field of the SGRIP feature extraction, the MentalStat