tooth arrangement collections

Sinthanayothin, C., & Tharanon, W. (2007). Tooth alignment of the dental cast using 3D thin plate spline. Advances in Computer Science and Technology ACST, 2-4.
- Steps
1. 模拟单个牙齿和标记，并通过集合变形插入牙模中
2. 在tooth alignment前后记录牙齿标记
3. 通过采用3D thin plate spline技术生产新牙模 > 同时使用到牙冠和牙根的数据，对一颗完整的牙齿进行重排，而mesh数据中仅有牙冠，因此不适用。
Kumar, Y., Janardan, R., & Larson, B. (2012). Automatic feature identification in dental meshes. Computer-Aided Design and Applications, 9(6), 747-769.

本文主要针对虚拟正畸中的关键步骤：特征提取(牙齿表面特征：尖(cusps)，grooves(凹槽)，边缘(incisal edges)，marginal ridges(边缘脊)，occlusal surface boundary(咬合面边界)
虚拟正畸中输入数据是对3D mesh分割后所得到的单颗牙齿的集合。
- 尖(cusps): Watershed Algorithm
- 其他特征: 基于曲率和二维截面聚类 > 这篇文章对牙齿表面的各个特征的特征讲的蛮详细的
Lian, C., Wang, L., Wu, T. H., Wang, F., Yap, P. T., Ko, C. C., & Shen, D. (2020). Deep multi-scale mesh feature learning for automated labeling of raw dental surfaces from 3D intraoral scanners. IEEE transactions on medical imaging, 39(7), 2440-2450. > Precisely labeling teeth on digitalized 3D dental surface models is the precondition for tooth position rearrangements in orthodontic treatment planning

基于深度学习 MeshSegNet, 具有开源代码，是专门针对牙齿的，将牙齿加上labal，是自动排牙所必备的一步
Lian, C., Wang, L., Wu, T. H., Liu, M., Durán, F., Ko, C. C., & Shen, D. (2019, October). MeshSNet: Deep multi-scale mesh feature learning for end-to-end tooth labeling on 3D dental surfaces. In International Conference on Medical Image Computing and Computer-Assisted Intervention (pp. 837-845). Springer, Cham.

和上面那篇差不多
Zhang, Z., Ong, S. H., Zhong, X., & Foong, K. W. (2016). Efficient 3D dental identification via signed feature histogram and learning keypoint detection. Pattern Recognition, 60, 189-204.

方法：The Signed Feature Histogram
需要输入训练数据后再检测
主要是为了识别出牙齿中的特征点，也是自动排牙所必备的一步
Wei, G., Cui, Z., Liu, Y., Chen, N., Chen, R., Li, G., & Wang, W. (2020, August). TANet: Towards Fully Automatic Tooth Arrangement. In European Conference on Computer Vision (pp. 481-497). Springer, Cham.

采用卷积神经网络
分割完后加上标签，输入多层感知机，进行位置重排
使用PointNet提取特征点