{"paper":{"title":"Tracing Like a Clinician: Anatomy-Guided Spatial Priors for Cephalometric Landmark Detection","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Translating orthodontists' tracing workflow into anatomy-guided spatial priors improves landmark detection accuracy and generalization across imaging devices.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Pallavi Mohanty, Sidhartha Mohapatra","submitted_at":"2026-05-05T04:33:45Z","abstract_excerpt":"Clinicians trace cephalometric radiographs by following a structured anatomical workflow -- yet no prior system explicitly encodes this into computation. We present a five-phase anatomy-guided pipeline producing confidence-weighted spatial priors that shape HRNet-W32 training. The system achieves 1.04 mm mean radial error on 25 landmarks across 1,502 radiographs from 7+ imaging devices -- comparable to HYATT-Net (1.05 mm on CEPHA29) via explicit anatomical priors rather than learned attention. A three-way ablation isolates the mechanism: anatomical priors maintain a 1% validation-to-test gap, "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"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 that removing anatomical priors destroys generalization: both models converge to ~1.03 mm on validation, but diverge to 1.94 vs. 1.04 mm on the test set.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the five-phase pipeline faithfully and without bias translates the clinician's structured workflow (soft tissue identification, region partitioning, contour tracing, geometric landmark definitions) into accurate confidence-weighted spatial priors that remain valid across diverse imaging devices and patient populations.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"An anatomy-guided five-phase initialization pipeline encodes clinical workflow as spatial priors for HRNet-W32, achieving 1.04 mm mean radial error on 25 cephalometric landmarks across 1502 radiographs and outperforming prior SOTA by 15.4% while demonstrating superior generalization in ablations.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Translating orthodontists' tracing workflow into anatomy-guided spatial priors improves landmark detection accuracy and generalization across imaging devices.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"841f6b24ed28aa1fb2843b19f1946cbef83dc09f40bd4d55b0b4cacafd075b21"},"source":{"id":"2605.03358","kind":"arxiv","version":2},"verdict":{"id":"9fdc625c-2b81-410e-a038-9a1a9cdf603e","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-08T01:28:26.786556Z","strongest_claim":"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 that removing anatomical priors destroys generalization: both models converge to ~1.03 mm on validation, but diverge to 1.94 vs. 1.04 mm on the test set.","one_line_summary":"An anatomy-guided five-phase initialization pipeline encodes clinical workflow as spatial priors for HRNet-W32, achieving 1.04 mm mean radial error on 25 cephalometric landmarks across 1502 radiographs and outperforming prior SOTA by 15.4% while demonstrating superior generalization in ablations.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the five-phase pipeline faithfully and without bias translates the clinician's structured workflow (soft tissue identification, region partitioning, contour tracing, geometric landmark definitions) into accurate confidence-weighted spatial priors that remain valid across diverse imaging devices and patient populations.","pith_extraction_headline":"Translating orthodontists' tracing workflow into anatomy-guided spatial priors improves landmark detection accuracy and generalization across imaging devices."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.03358/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T14:33:35.623409Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-20T01:31:21.283877Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T15:27:47.077939Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"0f8bf63d08651db591db3d77d77aba1f3e6513fede61e01216b22f5952498230"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}