Teeth3DS An Extended Benchmark for Intraoral 3D Scans Analysis

2025-04-26 0 0 7.24MB 10 页 10玖币
侵权投诉
Teeth3DS+: An Extended Benchmark for Intraoral 3D Scans Analysis
Achraf Ben-Hamadoua,b,
, Nour Neifara,b, Ahmed Rekika,b, Oussama Smaouic, Firas Bouzguendac,
Sergi Pujadesd, Edmond Boyerd, Edouard Ladroitc
aSMARTS Laboratory, Technopark of Sfax, Sakiet Ezzit 3021, Sfax, Tunisia
bDigital Research Center of Sfax, Technopark of Sfax, Sakiet Ezzit 3021, Sfax, Tunisia
cUdini, 37 BD Aristide Briand, 13100 Aix-En-Provence, France
dInria, Univ. Grenoble Alpes, CNRS, Grenoble INP, LJK, France
Abstract
Intraoral 3D scans analysis is a fundamental aspect of Computer-Aided Dentistry (CAD) systems,
playing a crucial role in various dental applications, including teeth segmentation, detection, labeling,
and dental landmark identification. Accurate analysis of 3D dental scans is essential for orthodon-
tic and prosthetic treatment planning, as it enables automated processing and reduces the need for
manual adjustments by dental professionals. However, developing robust automated tools for these
tasks remains a significant challenge due to the limited availability of high-quality public datasets and
benchmarks. This article introduces Teeth3DS+, the first comprehensive public benchmark designed
to advance the field of intraoral 3D scan analysis. Developed as part of the 3DTeethSeg 2022 and
3DTeethLand 2024 MICCAI challenges, Teeth3DS+ aims to drive research in teeth identification, seg-
mentation, labeling, 3D modeling, and dental landmarks identification. The dataset includes at least
1,800 intraoral scans (containing 23,999 annotated teeth) collected from 900 patients, covering both
upper and lower jaws separately. All data have been acquired and validated by experienced orthodon-
tists and dental surgeons with over five years of expertise. Detailed instructions for accessing the
dataset are available at https://crns-smartvision.github.io/teeth3ds
Keywords: Teeth3DS, Teeth3DS+, intraoral 3D scans, 3D point cloud, 3D segmentation, dentistry
1. Introduction
Computer-aided design (CAD) tools are becoming increasingly popular in modern dentistry for
highly accurate treatment planning. Advanced intra-oral scanners (IOSs) are particularly popular in or-
thodontic CAD software as they provide an accurate digital surface 3D representation of the dentition.
Such 3D representation can greatly assist dentists in simulating tooth extraction, realignment, and
smile design, making the treatment’s final results more predictable. As a result, digital teeth models
have the potential to relieve dentists from time-consuming and tedious tasks.
Although IOSs are becoming ubiquitous in clinical dental practice, there are only a few contri-
butions on teeth segmentation/labeling available in the literature, e.g., [1, 2, 3, 4, 5, 6], and, most
importantly, no publicly available benchmark. A fundamental challenge in IOS data analysis is the abil-
ity to accurately segment and identify teeth. Teeth segmentation and labeling are challenging due to
inter-class variations, such as inherent similarities between tooth shapes and ambiguous positions on
jaws, as well as intra-class variations such as damaged teeth or braces. This is in addition to the tight
Corresponding author
Email addresses: achraf.benhamadou@crns.rnrt.tn (Achraf Ben-Hamadou), (Nour Neifar), (Ahmed Rekik), (Oussama
Smaoui), (Firas Bouzguenda), (Sergi Pujades), (Edmond Boyer), (Edouard Ladroit)
Preprint submitted to Elsevier November 13, 2024
arXiv:2210.06094v2 [cs.CV] 11 Nov 2024
borders between surrounding teeth and gingiva as well as abnormalities like crowded, misaligned, or
missing teeth.
Despite several efforts, developing an accurate and a fully automated dental segmentation and
labeling tool remains challenging. Most former approaches are based on the teeth geometric char-
acteristics, such as curvature thresholding [7] or active contour computing [8] in order to separate
between teeth and gingiva. However, these solutions lack robustness with respect to teeth shape vari-
ations and usually require additional manual interventions that include teeth pre-selection or contour
initialization. Recent advances in 3D shape analysis based on machine learning techniques can also be
used for 3D dental scan analysis. In general, they follow two main strategies. Some researchers pro-
posed to convert the 3D scans into 2D images so that existing convolutional neural networks (CNNs)
can be used for both teeth segmentation and classification [9]. This strategy may lead to acceptable
results in the 2D space. Meanwhile, the reconstruction of the obtained segmentation into the 3D space
will systematically decrease the segmentation accuracy. Other studies proposed to directly process
the 3D scans and to apply deep learning architectures [10] with the aim of detecting and segmenting
teeth present in the IOSs. Although this approach may be faster, it still has to deal with different chal-
lenges such as point cloud irregularity, downsampling, and 3D pose variations. Overall, it should also
be noted that the lack of a publicly available dataset or benchmark is an obstacle to the development
of this research area, as no fair comparison can be conducted among state-of-the-art approaches.
In this article, we introduce a freely available teeth dataset of 1800 intra-oral scans collected from
900 patients covering the upper and lower jaws separately. These data were first used for the 3DTeeth-
Seg 2022 [11, 12] scientific challenge held during the 25th International Conference on Medical Image
Computing and Computer Assisted Intervention (MICCAI) (https://3dteethseg.grand-challenge.org/),
which reflects the research community’s interest in our proposed dataset. The provided data can be
used by researchers of different backgrounds for the development and evaluation of machine learning
methods not only for teeth detection, segmentation and labeling, but also other tasks such as 3D shape
modeling and reconstruction from 2D intra-oral photos.
2. Methods
2.1. Patients and intra-oral scans collection
In compliance with the European General Data Protection Regulation (GDPR) agreement, we ob-
tained 3D intra-oral scans for 900 patients acquired by orthodontists/dental surgeons with more than
5 years of professional experience from partner dental clinics located mainly in France and Belgium.
All data is completely anonymized, and the identity of the patients cannot be revealed. Two 3D scans
are acquired for each patient, covering the upper and lower jaws separately. The following IOSs were
used for scan acquisition: the Primescan from Dentsply, the Trios3 from 3Shape, and the iTero Element
2 Plus. These scanners are representative and generate 3D scans with an accuracy between 10 and
90 micrometers and a point resolution between 30 and 80 pts/mm2. No additional equipment other
than the IOS itself was used during the acquisitions. All acquired clinical data are collected for pa-
tients requiring either orthodontic (50%) or prosthetic treatment (50%). The provided dataset follows
a real-world patient age distribution: 50% male 50% female, about 70% under 16 years-old, about 27%
between 16-59 years-old, about 3% over 60 years old.
2.2. Data annotation and processing
2.2.1. Teeth detection, segmentation, and labeling
The data annotation, i.e., teeth segmentation and labeling, was performed in collaboration with clin-
ical evaluators with more than 10 years of expertise in orthodontistry, dental surgery, and endodontics.
The detailed process is depicted in Figure 1. It consists of eight steps. First, the 3D scans are pre-
processed (steps 1 and 2 in Figure 1) by removing all degenerated and redundant mesh faces, as well
2
摘要:

Teeth3DS+:AnExtendedBenchmarkforIntraoral3DScansAnalysisAchrafBen-Hamadoua,b,∗,NourNeifara,b,AhmedRekika,b,OussamaSmaouic,FirasBouzguendac,SergiPujadesd,EdmondBoyerd,EdouardLadroitcaSMARTSLaboratory,TechnoparkofSfax,SakietEzzit3021,Sfax,TunisiabDigitalResearchCenterofSfax,TechnoparkofSfax,SakietEzzi...

展开>> 收起<<
Teeth3DS An Extended Benchmark for Intraoral 3D Scans Analysis.pdf

共10页,预览2页

还剩页未读, 继续阅读

声明:本站为文档C2C交易模式,即用户上传的文档直接被用户下载,本站只是中间服务平台,本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私,请立即通知玖贝云文库,我们立即给予删除!
分类:图书资源 价格:10玖币 属性:10 页 大小:7.24MB 格式:PDF 时间:2025-04-26

开通VIP享超值会员特权

  • 多端同步记录
  • 高速下载文档
  • 免费文档工具
  • 分享文档赚钱
  • 每日登录抽奖
  • 优质衍生服务
/ 10
客服
关注