Skelite: Compact Neural Networks for Efficient Iterative Skeletonization

2025-03-10

AI Summary

Existing skeletonization algorithms face a fundamental trade-off between computational efficiency and topological fidelity: morphological methods are fast but prone to skeletal fragmentation, whereas topology-preserving approaches achieve high accuracy at prohibitive computational cost—particularly challenging for clinical applications such as vascular structure extraction from medical images. Method: We propose the first fully differentiable, lightweight iterative skeletonization framework tailored for curve-like anatomical structures (e.g., vasculature). Our approach integrates synthetic data training, task-aware augmentation, and knowledge distillation for end-to-end optimization, and supports unified 2D/3D processing with built-in post-processing. Contribution/Results: Our method accelerates state-of-the-art topology-preserving algorithms by 100× while significantly improving skeletal connectivity. It generalizes zero-shot across imaging centers and modalities without fine-tuning, and demonstrates strong efficacy and robustness in downstream tasks—including vessel segmentation—validating its clinical applicability.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Structured Learning
• Search and Optimization: Distributed Search

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Responsible Web: Algorithmic accountability and transparency on the web

Abstract

Skeletonization extracts thin representations from images that compactly encode their geometry and topology. These representations have become an important topological prior for preserving connectivity in curvilinear structures, aiding medical tasks like vessel segmentation. Existing compatible skeletonization algorithms face significant trade-offs: morphology-based approaches are computationally efficient but prone to frequent breakages, while topology-preserving methods require substantial computational resources. We propose a novel framework for training iterative skeletonization algorithms with a learnable component. The framework leverages synthetic data, task-specific augmentation, and a model distillation strategy to learn compact neural networks that produce thin, connected skeletons with a fully differentiable iterative algorithm. Our method demonstrates a 100 times speedup over topology-constrained algorithms while maintaining high accuracy and generalizing effectively to new domains without fine-tuning. Benchmarking and downstream validation in 2D and 3D tasks demonstrate its computational efficiency and real-world applicability

Problem

• Efficient skeletonization with minimal computational resources
• Preserving connectivity in curvilinear structures for medical tasks
• Balancing speed and accuracy in topology-preserving algorithms

Innovation

• Uses synthetic data and task-specific augmentation
• Implements model distillation for compact networks
• Achieves 100x speedup with high accuracy