The invention relates to the field of medical
image segmentation, and discloses a three-dimensional oral
hard palate image segmentation method based on a multidirectional
state space model. The method comprises the following steps: acquiring a target image
data set which comprises a
CBCT image data set and a corresponding sketching
label image
data set; constructing a multidirectional
state space model; taking the target image data set as a
training set, and training a multidirectional
state space model; and inputting a
CBCT image to be segmented into the multidirectional state
space model, and outputting a segmentation result of the three-dimensional oral
hard palate image. According to the method, the understanding ability of the model for a three-dimensional structure can be enhanced, the complex
spatial dependency relationship of the oral
hard palate region is effectively captured, the accuracy and robustness of the segmentation model are improved, and the calculation efficiency and precision of an existing segmentation method are improved.
Semantics and uncertainty are introduced in training to perform
hybrid data enhancement, the generalization and precision of the network can be improved, the sketching efficiency is improved, the sketching burden of a dentist is reduced, and a support is provided for realizing an efficient, accurate and generalized orthodontic planting plan.