The invention discloses a CT
aorta robustness three-dimensional segmentation method and device based on
deep learning, and relates to the technical field of medical
image processing, and the method comprises the steps: extracting an
aorta slice sequence frame by frame; carrying out linear interpolation and
thresholding on the extracted probability graph; setting the initial segmentation
mask of the current slice as an initialization prompt of a medical video segmentation model with a memory mechanism; for the initial slice image, calling an initialized memory
state function through a medical video segmentation model, generating an initial segmentation result of the initial slice, and encoding the initial segmentation result into an initial memory state; calling a propagation tracking
state function to automatically predict an
aorta original
mask of a current slice through a medical video segmentation model by using memory states established in previous several frames of slices and a current slice image, and obtaining a prediction
mask; and carrying out weighted fusion and three-dimensional reconstruction on the prediction mask and the preliminary segmentation mask. According to the method, full-automatic and robust three-dimensional segmentation of the CT aorta is realized.