Georgia Institute of Technology
AI-Driven 4D Lung SBRT Treatment Planning: An End-to-End Approach
Abstract
dc:description.abstractCurrently, patient motion management in 4D lung SBRT cases is primarily based on subjective judgment from the involved staff, namely the radiation oncologist. To obtain more objective information without adding significant cost overhead to the clinical workflow, this work improves registration and segmentation methods using deep learning techniques and utilizes built-in treatment planning software in Eclipse. Image registration and OAR segmentation are time-consuming and error-prone steps in the treatment planning workflow for external beam radiation. In this work, a registration method is developed to improve numerical results and deformation realism. The registration model uses cycle consistency to improve bi-directional registration consistency, a generator-discriminator pair to force realistic deformations, and weak supervision to improve registration quality. Additionally, segmentation models were created and compared to current industry-standard techniques and radiation oncologist-approved contours using Eclipse. Eclipse is further used to generate DVH curves to aid radiation oncologist decision making regarding motion management selection for 4DCT lung SBRT cases. These steps can be utilized in a single, mostly automated workflow to aid in phase gating selection.
Degree
thesis:*- Level thesis:degree_level
- Doctoral
- Department dc:contributor.department
- Mechanical Engineering
- Grantor dc:publisher
- Georgia Institute of Technology
- Year dc:date.issued
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Matkovic, Luke Alexander
- Advisors dc:contributor.advisor
-
- Wang, C.-K. Chris
- Yang, Xiaofeng
- Committee members dc:contributor.committeemember
-
- Elder, Eric
- Dewji, Shaheen
- Biegalski, Steven
Subjects
dc:subject × 6Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/1853/76867