Progressive Curriculum Learning with Scale-Enhanced U-Net for Continuous Airway Segmentation

2024-10-24

AI Summary

To address intra-class imbalance (e.g., pronounced disparities between main bronchi and fine peripheral branches) and structural discontinuities caused by blurred fine details in chest CT airway segmentation, this paper proposes a three-stage progressive curriculum learning framework integrating an Adaptive Topological Response Loss (ATRL) and a Scale-Enhanced U-Net (SE-UNet). The SE-UNet incorporates multi-scale inputs and a Detail-Intensified Enhancement (DIE) module to strengthen feature representation of subtle airways; ATRL explicitly enforces topological continuity of the airway tree. Extensive validation across multiple internal and external datasets demonstrates substantial improvements in recall of small airways and structural connectivity. Key metrics—including small-branch recall and airway-tree completeness—achieve state-of-the-art performance. The method delivers more robust airway parsing, thereby enhancing preoperative planning and bronchoscopic navigation.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Structured Learning
• Planning, Routing, and Scheduling: Activity and Plan Recognition

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Bridging structured and unstructured data
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Continuous and accurate segmentation of airways in chest CT images is essential for preoperative planning and real-time bronchoscopy navigation. Despite advances in deep learning for medical image segmentation, maintaining airway continuity remains a challenge, particularly due to intra-class imbalance between large and small branches and blurred CT scan details. To address these challenges, we propose a progressive curriculum learning pipeline and a Scale-Enhanced U-Net (SE-UNet) to enhance segmentation continuity. Specifically, our progressive curriculum learning pipeline consists of three stages: extracting main airways, identifying small airways, and repairing discontinuities. The cropping sampling strategy in each stage reduces feature interference between airways of different scales, effectively addressing the challenge of intra-class imbalance. In the third training stage, we present an Adaptive Topology-Responsive Loss (ATRL) to guide the network to focus on airway continuity. The progressive training pipeline shares the same SE-UNet, integrating multi-scale inputs and Detail Information Enhancers (DIEs) to enhance information flow and effectively capture the intricate details of small airways. Additionally, we propose a robust airway tree parsing method and hierarchical evaluation metrics to provide more clinically relevant and precise analysis. Experiments on both in-house and public datasets demonstrate that our method outperforms existing approaches, significantly improving the accuracy of small airways and the completeness of the airway tree. The code will be released upon publication.

Problem

• Enhance continuous airway segmentation in chest CT images
• Address intra-class imbalance between large and small airways
• Improve accuracy and completeness of small airway segmentation

Innovation

• Progressive curriculum learning pipeline
• Scale-Enhanced U-Net with multi-scale inputs
• Adaptive Topology-Responsive Loss for continuity