The invention discloses a method for segmenting
renal cortex images. The method comprises the following steps of: 1, performing initial segmentation on a
kidney image by using a statistical shape model
algorithm to acquire grid images of an initial outer surface S' of a
kidney structure; 2, establishing sub-graphs G0 and G1 for the outer surface and the inner surface of a
renal cortex respectively in a
narrow band close to the grid images of the initial outer surface S' so as to obtain a weighted
directed graph G=(G0, G1), designing cost functions C0 and C1 by aiming at the outer surface and the inner surface of the
renal cortex respectively, generating an s-t graph Gst based on the weighted
directed graph G=(G0, G1) and the cost functions C0 and C1, and extracting two
layers of optimal surface images, namely an outer surface image S0 and an inner surface image S1 of the renal cortex, by using a max-flow minimal
cut method; and 3, generating a data image Vcortex of a renal cortex body by using the outer surface image S0 and the inner surface image S1 of the renal cortex. By the method, the renal cortex structure can be segmented accurately, and the
workload for manually segmenting the renal cortex is greatly reduced, so the method has important application values in the field of segmenting the renal cortex images and
ventricular wall images with multi-surface structural characteristics.