Feature selection method and device, feature extraction method and device, anomaly detection model and construction method of anomaly detection model

The invention relates to a feature selection method and device, a feature extraction method and device, an anomaly detection model and a construction method thereof. The feature selection method comprises the following steps: acquiring a brain perfusion image set, wherein brain perfusion images in the brain perfusion image set comprise brain images at multiple moments; determining a region of interest of the brain perfusion image, wherein the region of interest comprises a normal region and an abnormal region; based on the plurality of moments, respectively extracting first image features of the region of interest; and based on a multi-level feature selection strategy, screening out a second image feature used for distinguishing the normal region and the abnormal region from the first image feature. According to the embodiment of the invention, the dynamic features can be effectively extracted, and the feature selection precision is improved. 本公开涉及一种特征选择、提取方法及装置、异常检测模型及其构建方法。所述特征选择方法包括获取脑灌注影像集，所述脑灌注影像集中的脑灌注影像