Surface-to-Skeleton 3D Cephalometry: Estimating Hidden Skeletal Landmarks from CT-Derived External Soft-Tissue Surfaces

2026-08-12

AI Summary

This study addresses the challenge of accurately inferring craniofacial skeletal landmarks—unobservable in CT—from the geometry of external soft tissue surfaces. Leveraging paired data of soft-tissue point clouds and skeletal landmarks derived from co-registered CT scans, the work formulates a coordinate-consistent surface-to-bone mapping task and introduces a hierarchical point cloud neural network for implicit landmark estimation. For the first time, the feasibility of this inference is rigorously validated under controlled confounding factors, including registration accuracy, scanning domain, and acquisition conditions, revealing the critical contribution of non-anterior geometric information. Evaluated on a hold-out cohort of 40 patients, the method achieves a mean radial error of 2.97 mm (3.03 mm for deep landmarks), demonstrating the presence of subject-specific signals that surpass population-average anatomical configurations.

Abstract

Existing 3D facial-landmark methods localize points on visible skin, but whether CT-defined internal skeletal landmarks can be inferred from external soft-tissue geometry remains unclear. We formulate a coordinate-consistent surface-to-skeleton task using same-acquisition CT-derived surfaces, separating estimation from optical-to-CT registration, scanner-domain, and acquisition-state effects, with coverage analyzed separately. From 240 clinical CT scans from two hospitals, we construct a locked retrospective protocol pairing CT-derived external soft-tissue point clouds with 21 skeletal landmarks and three visible soft-tissue landmarks. An integrated hierarchical point-cloud model achieves 2.97 mm mean radial error on skeletal landmarks and 3.03 mm on deep or surface-invisible landmarks in 40 held-out patients. Patient-mismatch controls support patient-specific signal beyond a fixed population configuration or global similarity alone, while coverage ablations indicate dependence on non-anterior geometry. Optical-transfer diagnostics reveal substantial coverage-related and global-configuration components, although deployable optical inference remains unresolved. These results answer the controlled feasibility question affirmatively and provide a basis for hidden skeletal landmark inference.

Problem

• 3D cephalometry
• skeletal landmarks
• soft-tissue surfaces
• CT scans
• surface-to-skeleton inference

Innovation

• surface-to-skeleton
• 3D cephalometry
• point-cloud model
• skeletal landmark estimation
• CT-derived soft-tissue