🤖 AI Summary
Existing automated methods struggle to emulate the reasoning process orthodontists employ when localizing cephalometric landmarks based on anatomical structures and geometric rules. This work proposes a five-stage anatomy-guided initialization pipeline that, for the first time, encodes clinical anatomical knowledge as spatial attention priors and integrates them into an HRNet-W32 detector via confidence-weighted fusion. This inductive bias substantially enhances the model’s generalization capability. Evaluated on 1,502 multi-source cephalograms, the method achieves a mean radial error of 1.04 mm across 25 landmarks—representing a 15.4% improvement over the current state of the art—with 12 landmarks exhibiting sub-millimeter accuracy. Ablation studies confirm that the performance gain stems from improved anatomical plausibility rather than the addition of extra input channels.
📝 Abstract
When orthodontists trace cephalometric radiographs, they follow a structured workflow: identify the soft tissue profile, partition the skull into anatomical regions, trace contours, and locate landmarks using geometric definitions -- yet no automated system replicates this reasoning. We present a five-phase anatomy-guided initialization pipeline that translates this clinical workflow into computational operations, producing confidence-weighted spatial attention priors for a downstream HRNet-W32 detector. On 1,502 radiographs from three sources spanning 7+ imaging devices, the system achieves 1.04 mm mean radial error on 25 landmarks -- surpassing prior state-of-the-art (1.23 mm on 19 landmarks) by 15.4%, with twelve landmarks below 1 mm. A three-way controlled ablation reveals two striking findings. First, removing anatomical priors does not merely slow convergence -- it destroys generalization: both models converge to ~1.03 mm on validation, but diverge to 1.94 vs. 1.04 mm on the test set. Second, replacing anatomical priors with random-position Gaussians produces even worse generalization (2.24 mm), confirming that the improvement derives from anatomically correct positioning, not additional input channels. Clinical domain knowledge encoded as spatial priors provides an inductive bias that architecture and data augmentation alone do not provide.