![[2512.07984] Restrictive Hierarchical Semantic Segmentation for Stratified Tooth Layer Detection](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2512.07984] Restrictive Hierarchical Semantic Segmentation for Stratified Tooth Layer Detection
Summary
This article presents a novel framework for hierarchical semantic segmentation aimed at improving the detection of stratified tooth layers, enhancing accuracy in dental imaging.
Why It Matters
Accurate segmentation of dental structures is crucial for diagnosing and staging dental diseases. This research introduces a method that leverages hierarchical relationships to improve segmentation performance, which could lead to better clinical outcomes in dental practices, especially in data-scarce environments.
Key Takeaways
- Introduces a framework that embeds anatomical hierarchy into semantic segmentation.
- Demonstrates improved performance in detecting fine-grained anatomical structures.
- Utilizes a novel hierarchical loss function to enhance prediction consistency.
- Validates the approach using a new dataset of panoramic radiographs.
- Findings suggest increased recall but also a rise in false positives.
Computer Science > Computer Vision and Pattern Recognition arXiv:2512.07984 (cs) [Submitted on 8 Dec 2025 (v1), last revised 19 Feb 2026 (this version, v4)] Title:Restrictive Hierarchical Semantic Segmentation for Stratified Tooth Layer Detection Authors:Ryan Banks, Camila Lindoni Azevedo, Hongying Tang, Yunpeng Li View a PDF of the paper titled Restrictive Hierarchical Semantic Segmentation for Stratified Tooth Layer Detection, by Ryan Banks and Camila Lindoni Azevedo and Hongying Tang and Yunpeng Li View PDF HTML (experimental) Abstract:Accurate understanding of anatomical structures is essential for reliably staging certain dental diseases. A way of introducing this within semantic segmentation models is by utilising hierarchy-aware methodologies. However, existing hierarchy-aware segmentation methods largely encode anatomical structure through the loss functions, providing weak and indirect supervision. We introduce a general framework that embeds an explicit anatomical hierarchy into semantic segmentation by coupling a recurrent, level-wise prediction scheme with restrictive output heads and top-down feature conditioning. At each depth of the class tree, the backbone is re-run on the original image concatenated with logits from the previous level. Child class features are conditioned using Feature-wise Linear Modulation of their parent class probabilities, to modulate child feature spaces for fine grained detection. A probabilistic composition rule enforces consistenc...