# 3DTeethSeg'22: 3D Teeth Scan Segmentation and Labeling Challenge

**Authors:** Achraf Ben-Hamadou, Oussama Smaoui, Ahmed Rekik, Sergi Pujades, Edmond Boyer, Hoyeon Lim, Minchang Kim, Minkyung Lee, Minyoung Chung, Yeong-Gil Shin, Mathieu Leclercq, Lucia Cevidanes, Juan Carlos Prieto, Shaojie Zhuang, Guangshun Wei, Zhiming Cui, Yuanfeng Zhou, Tudor Dascalu, Bulat Ibragimov, Tae-Hoon Yong  
**arXiv:** [2305.18277](https://arxiv.org/abs/2305.18277)  
**Date:** 2023-05-29  
**Summary page:** https://www.summarizepaper.com/en/paper/2305.18277v1/

## Key points

- Teeth localization, segmentation, and labeling from intra-oral 3D scans are crucial tasks in dentistry.
- The 3DTeethSeg'22 challenge was organized as part of the MICCAI conference in 2022.
- The challenge aimed to develop algorithms for accurately localizing, segmenting, and labeling teeth from intraoral 3D scans.
- A dataset containing 1800 scans from 900 patients was prepared with each tooth individually annotated using a human-machine hybrid algorithm.
- Six algorithms were evaluated based on their teeth segmentation accuracy (TSA) and teeth identification rate (TIR).
- Over five hundred requests for data download and registration were received for the challenge.
- Forty-four teams participated in the preliminary phase and ten teams submitted final submissions for evaluation.
- The top six ranked methods included an end-to-end network incorporating graph attentional convolution layers, a two-stream graph convolutional network (TSGCN), and Dental Arch (DArch).
- The dataset used in the challenge was acquired following GDPR agreement from partner dental clinics mainly located in Europe.

## Summary

Teeth localization, segmentation, and labeling from intra-oral 3D scans are crucial tasks in dentistry for improving dental diagnostics, treatment planning, and population-based studies on oral health. To address the challenges associated with developing automated algorithms for teeth analysis due to variations in dental anatomy, imaging protocols, and limited availability of publicly accessible data, the 3DTeethSeg'22 challenge was organized as part of the International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI) in 2022. The challenge aimed to develop algorithms that can accurately localize, segment, and label teeth from intraoral 3D scans. For this purpose, a dataset containing 1800 scans from 900 patients was prepared with each tooth individually annotated using a human-machine hybrid algorithm. Six algorithms were evaluated on this dataset based on their performance in terms of teeth segmentation accuracy (TSA) and teeth identification rate (TIR). Over five hundred requests for data download and registration were received for the challenge with forty-four teams participating in the preliminary phase and ten teams submitting their final submissions for evaluation. The top six ranked methods were selected based on their performance which included an end-to-end network incorporating graph attentional convolution layers and a global structure branch; a two-stream graph convolutional network (TSGCN); and Dental Arch (DArch), a method introduced for 3D tooth instance segmentation using weak low-cost annotated data. The dataset used in this challenge was acquired from orthodontists/dental surgeons with extensive professional experience following European General Data Protection Regulation (GDPR) agreement from partner dental clinics mainly located in Europe. In summary, the 3DTeethSeg'22 challenge aimed to advance automated teeth analysis by developing algorithms for teeth localization, segmentation, and labeling from intraoral 3D scans.