OsteoFlow: Lyapunov-Guided Flow Distillation for Predicting Bone Remodeling after Mandibular Reconstruction

2026-03-23

AI Summary

This study addresses the challenges of inconsistent remodeling trajectories and anatomical distortion in long-term mandibular bone remodeling prediction following reconstructive surgery. The authors propose a flow-based predictive framework that accurately forecasts one-year postoperative bone morphology using only a CT scan acquired on postoperative day 5. The core innovation lies in a Lyapunov stability–guided trajectory distillation mechanism, which learns continuous-time deformation trajectories from a velocity field derived via differentiable registration in a teacher model. To preserve geometric correspondence, the method further incorporates a resection-aware image loss tailored to the surgical defect region. Experimental results on 344 paired regions of interest demonstrate significant improvements over existing approaches, achieving approximately a 20% reduction in mean absolute error within the surgically resected areas.

Technology Category

• Machine Learning: Feature Construction/Reformulation
• Planning, Routing, and Scheduling: Replanning and Plan Repair
• Computer Vision: Motion & Tracking

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Predicting long-term bone remodeling after mandibular reconstruction would be of great clinical benefit, yet standard generative models struggle to maintain trajectory-level consistency and anatomical fidelity over long horizons. We introduce OsteoFlow, a flow-based framework predicting Year-1 post-operative CT scans from Day-5 scans. Our core contribution is Lyapunov-guided trajectory distillation: Unlike one-step distillation, our method distills a continuous trajectory over transport time from a registration-derived stationary velocity field teacher. Combined with a resection-aware image loss, this enforces geometric correspondence without sacrificing generative capacity. Evaluated on 344 paired regions of interest, OsteoFlow significantly outperforms state of-the-art baselines, reducing mean absolute error in the surgical resection zone by ~20%. This highlights the promise of trajectory distillation for long-term prediction. Code is available on GitHub: OsteoFlow.

Problem

• bone remodeling
• mandibular reconstruction
• long-term prediction
• trajectory consistency
• anatomical fidelity

Innovation

• Lyapunov-guided trajectory distillation
• flow-based generative model
• bone remodeling prediction
• resection-aware image loss
• continuous trajectory distillation