3D Functional-Anatomical Feature Matching Across Follow-Up Scans

The patent segments the image matching problem into multiple functional-anatomical feature layers (e.g., functional features like tracer uptake and anatomical features like tissue structure). Each layer is processed independently to generate separate matching maps, which are then integrated to produce the final matching result. This segmentation allows the system to handle different types of features that may change differently over time, improving overall matching accuracy despite significant temporal changes in the scans.

The patent applies local quality by generating separate 3D feature matching maps for different functional and anatomical feature types at local regions of interest. Instead of applying a uniform matching approach across the entire image, the system tailors the matching process to specific feature layers (e.g., lesion regions vs. surrounding tissue), allowing optimal matching strategies to be applied locally based on the specific characteristics of each feature type and region.

The patent extracts and removes the problematic image alignment step from the feature propagation process. Instead of aligning entire images (which introduces distortions due to patient positioning changes), the system directly propagates features by comparing local feature patterns across scans using the multi-layer matching approach. This extraction of the alignment step eliminates the source of distortion while preserving the ability to propagate features accurately.

The patent introduces 3D feature matching maps as intermediary structures that facilitate feature propagation without requiring direct image alignment. These matching maps serve as mediators that capture the spatial relationships and feature correspondences between scans, allowing features to be propagated through the intermediary maps rather than through direct geometric transformation of the original images, thus avoiding alignment-induced distortions.