Off-campus WSU users: To download campus access dissertations, please use the following link to log into our proxy server with your WSU access ID and password, then click the "Off-campus Download" button below.
Non-WSU users: Please talk to your librarian about requesting this thesis through interlibrary loan.
Access Type
WSU Access
Date of Award
January 2024
Degree Type
Thesis
Degree Name
M.S.
Department
Computer Science
First Advisor
Dongxiao Zhu
Abstract
Deep learning has transformed the landscape of various scientific fields, including computer vision, natural language processing, and medical image analysis, with both convolutional neural networks (CNNs) and transformers emerging as a cornerstone in healthcare informatics and biomedical imaging for tasks such as image classification and segmentation. Particularly, U-shaped networks have shown remarkable efficacy in medical image segmentation, crucial for enhancing the precision of medical diagnostics and treatments by extracting significant details from intricate image data.
In this thesis, we introduced a novel architecture termed SwinAttUNet, which integrates convolutional and transformer-based networks. The thesis focuses on the automatic 3D segmentation of multiple organs in CT images, enhancing accuracy and efficiency in medical diagnostics and treatment planning. The SwinAttUNet architecture benefits from a dual-branch encoder combining convolutional and Swin Transformer elements, and an attention-enhanced decoder, ensuring precise and detailed organ segmentation. This hybrid approach leverages both local and global image features, significantly outperforming existing segmentation methods. The study validates the model's effectiveness through extensive testing on diverse datasets, showcasing its potential to transform clinical workflows and improve patient outcomes.
Recommended Citation
Li, Chengyin, "Leveraging Convolution And Self-Attention Mechanisms For Enhanced Feature Extraction In Medical Image Segmentation" (2024). Wayne State University Theses. 951.
https://digitalcommons.wayne.edu/oa_theses/951