Abstract
Airway segmentation is critical for virtual bronchoscopy and computer-aided pulmonary disease analysis. In recent years, convolutional neural networks (CNNs) have been widely used to delineate the bronchial tree. However, the segmentation results of the CNN-based methods usually include many discontinuous branches, which need manual repair in clinical use. A major reason for the breakages is that the appearance of the airway wall can be affected by the lung disease as well as the adjacency of the vessels, while the network tends to overfit to these special patterns in the training set. To learn robust features for these areas, we design a multi-branch framework that adopts the geodesic distance transform to capture the intensity changes between airway lumen and wall. Another reason for the breakages is the intra-class imbalance. Since the volume of the peripheral bronchi may be much smaller than the large branches in an input patch, the common segmentation loss is not sensitive to the breakages among the distal branches. Therefore, in this paper, a breakage-sensitive regularization term is designed and can be easily combined with other loss functions. Extensive experiments are conducted on publicly available datasets. Compared with state-of-the-art methods, our framework can detect more branches while maintaining competitive segmentation performance.
Introduction: The Challenge of Airway Breakages
Mapping the bronchial tree from CT scans is a critical first step for virtual bronchoscopy, surgical navigation, and diagnosing lung diseases. While Convolutional Neural Networks (CNNs) are the standard tool for this 3D segmentation task, their outputs often suffer from "breakages"—discontinuous branches that render the 3D model topologically incorrect.
These breaks typically stem from three primary issues:
- Intra-class imbalance: Peripheral bronchi are microscopic compared to the main airways. Because standard loss functions are volume-based, a network isn't heavily penalized for missing these tiny distal branches.
- Vessel interference: Nearby blood vessels can have similar or higher densities (Hounsfield units) than airway walls, confusing the model's boundary detection.
- Pathology: Lung diseases and lesions can deform the airway walls, creating irregular shapes and thicknesses that the network struggles to recognize.
Methodology: The BREAK Framework
To address these challenges, the authors propose a two-stage, multi-branch architecture known as BREAK.
Geodesic Distance Transform (GDT)
While airway walls change appearance due to disease or nearby vessels, the lumen (the hollow inside of the airway) remains consistently dark. The framework uses GDT to calculate the shortest path from centerline voxels to other voxels based on image intensity. This effectively traces the dark tubes of the airway, ignoring deformed walls and providing a more robust feature set for segmentation.
Breakage-Sensitive (BS) Loss
Standard loss functions like Dice or Cross-Entropy are volume-centric. Missing a tiny branch barely affects the total loss. BS loss addresses this by calculating loss based specifically on the airway's centerline (its topological skeleton). This applies a significant penalty if a branch breaks, regardless of the branch's volume, forcing the network to prioritize connectivity.
Skeleton Embedding Fusion
The framework trains two branches in parallel: one fine-tuned for high precision and another using GDT for high continuity. A fusion algorithm then takes the unbroken centerlines from the GDT branch and embeds them into the high-precision output, effectively patching any gaps in the final prediction.
Key Insights and Results
The proposed method achieved state-of-the-art performance, particularly in its ability to detect more of the airway tree without fracturing. On the public Binary Airway Segmentation Dataset (BASD), it achieved a 95.33% tree length detected rate and a 94.22% branch detected rate, outperforming baseline methods by 2% to 5%.
Topology over Volume
By shifting the loss function's focus from total volume to the 1D centerline (BS loss), the network is forced to care about the connectivity of the smallest peripheral branches just as much as the main trachea. Ablation studies in the original paper showed that the BS loss alone is vastly more sensitive to structural fractures than standard losses, reacting sharply the moment a branch breaks during training.
The "False Positive" Paradox
Interestingly, the model became so proficient at finding tiny, faint airways that it detected branches human expert annotators had missed. In standard evaluation metrics, these are penalized as "false positives," even though retrospective clinical review confirmed they were actual, true airways.
Limitations and Future Work
Despite its success, the model faces a precision vs. sensitivity trade-off. The hyper-focus on finding tiny branches via BS loss can occasionally cause slight leakage into background textures. Furthermore, while GDT is robust against wall deformation, calculating it using image gradients can be sensitive to extreme noise. Future work will likely involve developing AI-assisted ground-truth generation to better evaluate these highly sensitive models against human-annotated datasets.
Summary updated