Tianran Yuan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Tianran Yuan has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
- Jiaji Liu
Guilin University of Electronic Technology, Guilin, China
1 shared publication - Jianzhuang Zhao
Human-Robot Interfaces and Interaction Lab, Italian Institute of Technology, Genova, Italy
1 shared publication Minquan Long
Guilin University of Electronic Technology, Guilin, China
1 shared publication- Nan Ye
Nanjing University of Aeronautics and Astronautics
1 shared publication Peng Tang
Technical University of Munich, School of Computation, Information, and Technology, Department of Informatics, Garching, Germany
1 shared publication- Shifeng Shi1 shared publication
Latest Publications
- 17 February 2026
- Published by Elsevier in Robotics and Computer-Integrated Manufacturing
- Vol. 100
3DVAC-HiPRoM: 3D vision and alignment control-based high-precise robot manipulation for manufacturing resilience
- 17 February 2026
- Published by Elsevier in Robotics and Computer-Integrated Manufacturing
- Vol. 100
3DVAC-HiPRoM: 3D vision and alignment control-based high-precise robot manipulation for manufacturing resilience
- 1 September 2025
- Published by SPIE-Intl Soc Optical Eng in Journal of Electronic Imaging
- Vol. 34 (05) , 053038-19
MMJTIS-Net: multi-model joint tooth instance segmentation network for advanced stomatological treatment
Automatic and accurate segmentation of individual teeth from cone beam computed tomography (CBCT) images is crucial for advanced stomatological treatments such as 3D-guided implant surgery and orthodontic planning. However, segmenting teeth from CBCT data poses significant challenges due to their proximity, poorly contrasted borders, and the presence of metal artifacts from prior treatments. Existing methods often suffer from overdependence on the initial detection step and difficulties in segmenting boundary and artifact regions. To address these issues, we propose MMJTIS-Net, a multi-model joint instance segmentation network that comprises a tooth instance detection component and a self-refined semantic segmentation component. The detection component incorporates an instance reference segmentation network to mitigate missed detections. The segmentation component employs a coarse-to-fine self-refined network that enhances boundary perception and addresses nonboundary difficult regions using a transformer-based self-attention mechanism. Results demonstrate that our approach achieves state-of-the-art performance, with a Dice similarity coefficient of 98.18% and a Jaccard index of 96.44%. In conclusion, our method not only enhances segmentation accuracy but also facilitates diverse clinical applications by delivering comprehensive 3D tooth information.
- 1 September 2025
- Published by SPIE-Intl Soc Optical Eng in Journal of Electronic Imaging
- Vol. 34 (05) , 053038-19
MMJTIS-Net: multi-model joint tooth instance segmentation network for advanced stomatological treatment
Automatic and accurate segmentation of individual teeth from cone beam computed tomography (CBCT) images is crucial for advanced stomatological treatments such as 3D-guided implant surgery and orthodontic planning. However, segmenting teeth from CBCT data poses significant challenges due to their proximity, poorly contrasted borders, and the presence of metal artifacts from prior treatments. Existing methods often suffer from overdependence on the initial detection step and difficulties in segmenting boundary and artifact regions. To address these issues, we propose MMJTIS-Net, a multi-model joint instance segmentation network that comprises a tooth instance detection component and a self-refined semantic segmentation component. The detection component incorporates an instance reference segmentation network to mitigate missed detections. The segmentation component employs a coarse-to-fine self-refined network that enhances boundary perception and addresses nonboundary difficult regions using a transformer-based self-attention mechanism. Results demonstrate that our approach achieves state-of-the-art performance, with a Dice similarity coefficient of 98.18% and a Jaccard index of 96.44%. In conclusion, our method not only enhances segmentation accuracy but also facilitates diverse clinical applications by delivering comprehensive 3D tooth information.