Cerebral hemorrhage image segmentation method based on attention mechanism and MSENet model

The invention discloses a cerebral hemorrhage image segmentation method based on an attention mechanism and an MSENet model. The method comprises the following steps: S1, acquiring a clinical cerebral hemorrhage CT image; s2, processing the clinical cerebral hemorrhage CT image through a multi-scale morphological processing module and outputting an anatomical radiomics feature map; s3, constructing an attention mechanism module, and performing training based on the anatomical radiomics feature map to obtain a trained attention mechanism module; and S4, constructing an MSENet network model, adding a plurality of trained attention mechanism modules to obtain a new MSENet network model, training the new MSENet network model to obtain a trained MSENet network model, and segmenting a cerebral hemorrhage area image from the clinical cerebral hemorrhage CT image. According to the method, the new MSENet network model added with the attention mechanism module is established, and the extracted feature map is calibrated