Respiration feature extraction method based on body surface significance analysis

A respiration feature extraction method based on body surface significance analysis, comprising:

• establishing a body surface voxel model of chest and abdomen respiration motion, which comprises acquisition of point cloud information, generation of the voxel model and extraction of respiration motion features;
• establishing a significance evaluation function by performing significance analysis on different regions of a body surface, and selecting the body surface region with high correlation with tumor motion based on the evaluation function;
• and
• performing voxelization on the body surface region, and obtaining effective one-dimensional characterization information of the body surface region using a local linear embedding dimension reduction algorithm,
• wherein the establishing a significance evaluation function by performing significance analysis on different regions of a body surface, and selecting the body surface region with high correlation with tumor motion based on the evaluation function comprises:
• performing equal-size region segmentation on the chest and abdomen body surface after three-dimensional reconstruction;
• performing significance analysis on each region of the body surface to obtain three significance characterization indexes of periodicity, stability and motion amplitude;
• performing dimension reduction on the three indexes using a kernel principal component analysis algorithm, so as to calculate a significance characterization value of each region;
• and
• selecting the region with the characterization value ranked at the top as an effective region.