IP3 2022 – Image Processing 2 (LOT 13763)
This lot is generally related to a technique to automatically segment digital image data which is useful in medical imaging. Disclosed is a technique to segment target structures in three-dimensional (3D) image data by pre-segmenting training images of structures representing the target structures and generating a model having inter-structure ordered spatial dependency for each of the target structures by placing all pre-segmented training images using neighboring structures as reference anchor structures. Also disclosed are techniques to integrate the ordered spatial dependency with a segmentation algorithm and sequentially segmenting each of the target structures based on the optimal sequence. The technology may be implemented in automatic medical image segmentation systems, smart image segmentation systems, interactive 3D image segmentation systems, etc.