Paravertebral muscle segmentation for body composition analysis in CT scans
In this study, deep learning techniques, specifically U-Net-based architectures, are used to automate the segmentation of paraspinal muscles in CT scans. This undergraduate research thesis aims to strengthen segmentation accuracy and improve body composition analysis (BCA) precision. Several models...
- Autores:
-
Gómez Mesa, Lina María
- Tipo de recurso:
- Trabajo de grado de pregrado
- Fecha de publicación:
- 2025
- Institución:
- Universidad de los Andes
- Repositorio:
- Séneca: repositorio Uniandes
- Idioma:
- eng
- OAI Identifier:
- oai:repositorio.uniandes.edu.co:1992/75992
- Acceso en línea:
- https://hdl.handle.net/1992/75992
- Palabra clave:
- Body composition analysis
Image processing
Paraspinal muscle
Semantic segmentation
Ingeniería
- Rights
- openAccess
- License
- Attribution 4.0 International
| Summary: | In this study, deep learning techniques, specifically U-Net-based architectures, are used to automate the segmentation of paraspinal muscles in CT scans. This undergraduate research thesis aims to strengthen segmentation accuracy and improve body composition analysis (BCA) precision. Several models are evaluated, including U-Net, U-Net++, AttU-Net, TransUNet, UNETR, and SwinUNETR to compare their performance under a transfer learning approach on the CAVAAT dataset. The models are evaluated using performance metrics such as the Dice Similarity Coefficient (DSC) and Hausdorff Distance (HD). In addition, SAM (Segment Anything Model) is tested for its adaptability to medical imaging tasks, providing insights into its accuracy and potential applications in scenarios with limited annotations. \href{https://github.com/Lina-go/PMSegmentation.git}{Github Repository} has models, data, and code. |
|---|