Abstract: A computer-implemented method and system for predicting orthodontic results based on landmark detection includes: acquiring a crown point cloud, a set of tooth point clouds, and tooth labels; extracting global dentition features and local tooth features, performing feature fusion on the global dentition features, tooth labels of individual teeth, and the local tooth features of the individual teeth to obtain fused features of the individual teeth, and extracting landmarks of the individual teeth and offset vectors from points in the tooth point clouds to the landmarks; fusing the landmarks of the individual teeth and the tooth point clouds, extracting tooth attention features, acquiring dentition attention features, and fusing the dentition attention features, the global dentition features, and the local tooth features to obtain a point cloud with fused landmarks; and acquiring pre- and post-treatment rigid transformation parameters, and obtaining a post-treatment crown model prediction result.

Abstract: A computer-implemented method and system for predicting orthodontic results based on landmark detection includes: acquiring a crown point cloud, a set of tooth point clouds, and tooth labels; extracting global dentition features and local tooth features, performing feature fusion on the global dentition features, tooth labels of individual teeth, and the local tooth features of the individual teeth to obtain fused features of the individual teeth, and extracting landmarks of the individual teeth and offset vectors from points in the tooth point clouds to the landmarks; fusing the landmarks of the individual teeth and the tooth point clouds, extracting tooth attention features, acquiring dentition attention features, and fusing the dentition attention features, the global dentition features, and the local tooth features to obtain a point cloud with fused landmarks; and acquiring pre- and post-treatment rigid transformation parameters, and obtaining a post-treatment crown model prediction result.