# SkullEngine: A Multi-stage CNN Framework for Collaborative CBCT Image Segmentation and Landmark Detection

**Authors:** Qin Liu, Han Deng, Chunfeng Lian, Xiaoyang Chen, Deqiang Xiao, Lei Ma, Xu Chen, Tianshu Kuang, Jaime Gateno, Pew-Thian Yap, James J. Xia  
**arXiv:** [2110.03828](https://arxiv.org/abs/2110.03828)  
**Date:** 2021-10-07  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2110.03828v1/

## Key points

- SkullEngine is a cutting-edge multi-stage CNN framework for collaborative CBCT image segmentation and landmark detection
- Utilizes a coarse-to-fine approach for high-resolution segmentation and large-scale landmark detection
- Key components include a collaborative JSD model and three refinement models
- Extensive evaluation conducted using a clinical dataset of 170 CBCT/CT images
- Primary objective: accurately segment midface and mandible bones, detect 175 clinically common landmarks on bones, teeth, and soft tissues
- Results demonstrate effectiveness and scalability in identifying landmarks and segmenting complex anatomical structures
- Represents significant advancement in medical image analysis for high-resolution segmentation and landmark detection in CBCT images
- Potential to enhance diagnostic accuracy and treatment planning in clinical settings

## Summary

SkullEngine is a cutting-edge multi-stage CNN framework developed by a team of researchers for collaborative CBCT image segmentation and landmark detection. The innovative framework utilizes a coarse-to-fine approach to achieve high-resolution segmentation and large-scale landmark detection. Its key components include a collaborative JSD model and three refinement models for segmentation and landmark detection. The team conducted an extensive evaluation of SkullEngine using a clinical dataset comprising 170 CBCT/CT images. Their primary objective was to accurately segment two crucial bones - the midface and mandible - while also detecting 175 clinically common landmarks on bones, teeth, and soft tissues. The results demonstrated the effectiveness and scalability of SkullEngine in accurately identifying landmarks and segmenting complex anatomical structures within medical imaging data. Overall, SkullEngine represents a significant advancement in the field of medical image analysis by providing a comprehensive solution for high-resolution segmentation and landmark detection in CBCT images. Its collaborative nature combined with its integrated JSD model showcases its potential for enhancing diagnostic accuracy and treatment planning in clinical settings.