Zygomaticomaxillary Suture (ZMS) dataset
收藏资源简介:
本数据集为全球首个公开的颧颌缝评估专用CBCT影像数据集,由深圳大学与昆明医科大学附属口腔医院联合创建,旨在解决正畸治疗中颧颌缝成熟度精准分期的临床难题。该数据集包含3,790个经专业标注的颧颌缝区域影像,数据源自2018年至2024年间约3,000名4-24岁患者的锥形束CT扫描,每个样本均按五阶段成熟度标准(A-E)进行独立标注,标注者间加权Kappa值达0.865,具有极高临床一致性。数据创建过程遵循严格临床协议,通过四位资深正畸医师手动截取颧颌缝区域并统一标注,在排除系统性疾病等干扰因素后,按患者级别进行8:2的训练-测试划分。该数据集主要应用于口腔正畸学领域,为颧颌缝成熟度自动评估算法开发提供关键数据支撑,助力解决因缝线高频细微变化和阶段间语义模糊导致的诊断标准化难题。
This is the world's first publicly available Cone Beam Computed Tomography (CBCT) imaging dataset specifically designed for zygomaticomaxillary suture assessment, jointly developed by Shenzhen University and the Affiliated Stomatological Hospital of Kunming Medical University. This dataset aims to address the clinical challenge of accurately staging the maturity of the zygomaticomaxillary suture during orthodontic treatment. It contains 3,790 professionally annotated images of the zygomaticomaxillary suture region, sourced from CBCT scans of approximately 3,000 patients aged 4 to 24 years between 2018 and 2024. Each sample was independently annotated in accordance with the five-stage maturity standard (A-E), with an inter-rater weighted Kappa coefficient of 0.865, demonstrating extremely high clinical consistency. The dataset was constructed following strict clinical protocols: four senior orthodontists manually cropped the regions of the zygomaticomaxillary suture and conducted unified annotations. After excluding confounding factors such as systemic diseases, the dataset was split into training and test subsets at an 8:2 ratio based on patient-level grouping. This dataset is primarily utilized in the field of orthodontics, providing critical data support for the development of automated zygomaticomaxillary suture maturity assessment algorithms, and facilitating the resolution of diagnostic standardization challenges caused by the high-frequency subtle changes of sutures and ambiguous semantic boundaries between adjacent stages.
数据集概述:SKMamba - 颧上颌缝评估数据集
数据集名称
- SKMamba(Structure-aware Knowledge-guided Heterogeneous Mamba)
任务目标
- 基于CBCT(锥形束CT)图像进行颧上颌缝成熟度分期的自动化评估。
所属会议/发表状态
- 论文已被 MICCAI 2026 接收。
数据集内容与结构
该数据集来源于Google Drive,文件包括:
- ZMS_classification.zip:主数据集压缩包,解压后包含五个类别文件夹:
A/B/C/D/E/
- descriptions.csv:文本描述文件。
- zms_text_features.npy:预提取的文本特征文件(NumPy格式)。
数据集的目录存放要求
ZMS_classification/解压后置于SKMamba/data/下descriptions.csv和zms_text_features.npy置于SKMamba/features/下
最终项目结构示例:
SKMamba/ ├── data/ │ └── ZMS_classification/ │ ├── A/ │ ├── B/ │ ├── C/ │ ├── D/ │ └── E/ ├── features/ │ ├── descriptions.csv │ └── zms_text_features.npy
数据划分文件(splits/)
- 用户需自行准备划分文件(如
splits/目录下的文本文件),每行包含图像相对路径,例如:
A/0012 LL.bmp A/0012 LU.bmp A/0012 RL.bmp A/0012 RU.bmp
标注信息
- 图像按类别(A、B、C、D、E)组织文件夹,类别标签即缝成熟度分期。
数据集获取方式
- 通过Google Drive下载:数据集与文本特征下载链接
许可协议
- 本项目采用 MIT License 开源协议。




