The application provides a multi-
modal head
image registration method based on stable markers, aiming at solving the problems of insufficient spatial alignment accuracy between multi-
modal medical images and poor robustness of traditional registration methods in cross-
modal scenarios. The method adopts a multi-anatomical marker
cascade strategy, realizes accurate registration from local to global through dual stable
anatomical structures of
ventricle and
skull. The method covers S1. acquiring a multi-modal head image dataset; S2. preprocessing the image data; S3. extracting the septum pellucidum cavity and the frontal horn of the lateral
ventricle in the head
ventricle as the first level marker points; S4. local registration based on the ventricle marker points, obtaining a primary
transformation matrix and applying it to the global image; S5. extracting the frontal pole, occipital pole and longitudinal
fissure of the brain in the head
skull as the second level marker points; S6. local registration based on the
skull marker points, obtaining a refined
transformation matrix and applying it to the global image; S7. calculating the registration accuracy evaluation index to verify the registration effect. The application uses two types of
anatomical structures with stable morphology and
high contrast in cross-modal imaging, i.e. ventricle and skull, as the registration reference, and gradually optimizes the
spatial transformation through the
cascade registration strategy, which significantly improves the registration accuracy and robustness of multi-modal head images, and provides a reliable spatial alignment basis for subsequent
clinical diagnosis, treatment planning and
image analysis.