Clustering-based shape segmentation for surface unfolding, model retrieval, and 3D printing applied in bolus shaping
| dc.contributor.author | Li, Rui | |
| dc.contributor.examiningcommittee | Xing, Malcolm (Mechanical Engineering) | en_US |
| dc.contributor.examiningcommittee | Leung, Carson (Computer Science) | en_US |
| dc.contributor.examiningcommittee | Tutunea-Fatan, O. Remus (Mechanical and Materials Engineering, Western University) | en_US |
| dc.contributor.supervisor | Peng, Qingjin | |
| dc.date.accessioned | 2022-03-15T19:05:39Z | |
| dc.date.available | 2022-03-15T19:05:39Z | |
| dc.date.copyright | 2022-03-15 | |
| dc.date.issued | 2022-03-14 | |
| dc.date.submitted | 2022-03-14T19:53:12Z | en_US |
| dc.date.submitted | 2022-03-15T18:44:35Z | en_US |
| dc.degree.discipline | Mechanical Engineering | en_US |
| dc.degree.level | Doctor of Philosophy (Ph.D.) | en_US |
| dc.description.abstract | Bolus is used to cover treatment areas of the patient skin to improve the dose distribution. The existing method of bolus shaping is a manual process to cut the material into pieces, then wrap and cover them on the target area, which is time-consuming and inaccurate. This research proposes clustering-based surface segmentation methods to improve accuracy and efficiency of the bolus shaping. Segmentation methods are developed for accurate shape unfolding, efficient model retrieval, and optimal 3D printing direction. For the accurate shape unfolding, surface flattenability is improved using a spectral clustering method based on features combined topology and segmentation saliency. A 3D mesh surface is optimized using a particle swarm optimizer with four requirements and one constraint to further improve surface unfolding accuracy. The 3D mesh surface is unfolded into 2D planes in a coordinate transformation process to improve the iterative efficiency using a mass-spring model with crossed springs. For the efficient model retrieval, an image-based method is proposed using visual entropies and feature skeletons of images and models. For the optimal 3D printing direction, an improved spectral clustering method is developed by combining the surface topology and printing features to improve surface quality and reduce support material. High-level features of the topology and printing are abstracted by Stacked Auto-encoders. Case studies verify the proposed segmentation and optimization methods. The shape unfolding method significantly reduces processing errors. The image-based model retrieval method shows its accuracy and efficiency in different applications. The surface segmentation increases surface quality and reduces support material of the 3D printed product. | en_US |
| dc.description.note | May 2022 | en_US |
| dc.identifier.uri | http://hdl.handle.net/1993/36366 | |
| dc.language.iso | eng | en_US |
| dc.rights | open access | en_US |
| dc.subject | Clustering | en_US |
| dc.subject | Shape segmentation | en_US |
| dc.subject | Bolus shaping | en_US |
| dc.title | Clustering-based shape segmentation for surface unfolding, model retrieval, and 3D printing applied in bolus shaping | en_US |
| dc.type | doctoral thesis | en_US |
| oaire.citation.endPage | 992 | en_US |
| oaire.citation.issue | 5 | en_US |
| oaire.citation.startPage | 979 | en_US |
| oaire.citation.title | Computer-aided design and application | en_US |
| oaire.citation.volume | 17 | en_US |
| project.funder.name | Discovery Grants from the Natural Sciences and Engineering Research Council (NSERC) of Canada | en_US |