3D multimodal image analysis for lung-cancer assessment

Open Access
- Author:
- Cheirsilp, Ronnarit
- Graduate Program:
- Computer Science
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- April 26, 2016
- Committee Members:
- William Evan Higgins, Dissertation Advisor/Co-Advisor
William Evan Higgins, Committee Chair/Co-Chair
Jesse Louis Barlow, Committee Member
Robert T Collins, Committee Member
Vishal Monga, Committee Member - Keywords:
- PET/CT
deformable registration
thoracic cavity
image processing
medical imaging
lung cancer
bronchoscopy - Abstract:
- Integrated positron emission tomography (PET) / computed-tomography (CT) scanners give 3D multimodal data sets that provide both molecular activity and anatomical information. Such data sets offer better identification of suspicious cancerous lesions and lymph nodes. Unfortunately, these PET/CT strengths have not yet been fully exploited by any currently available image-guided intervention (IGI) systems. The goal of this thesis is to develop a complete multimodal IGI system that enables effective and efficient use of PET/CT image data in the planning and guidance of bronchoscopy. To realize this, we have developed automatic and robust methods for segmenting the thoracic cavity and constructing a comprehensive chest model for follow-on optimal procedure planning, focused central-chest registration, data visualization, and computer-based bronchoscopy guidance. Subsequently, we proposed a deformable registration framework that enables the fusion of free-breathing whole-body PET/CT image data with breath-hold chest CT image data. The framework truly enables incorporation of important diagnostic PET information into the planning and guidance processes arising in cancer-staging bronchoscopy. Next, we developed PET/CT fusion, visualization, and detection methods that enable comprehensive interactive PET/CT visualization and PET-avid ROI selection. With the chest model, the methods provide unprecedented clarity of the central-chest lesions and neighboring anatomical structures truly relevant to lung-cancer assessment and follow-on procedure planning. We next devised computer-based tools and a unique complete multimodal system that integrate all methods summarized above. The tools provide PET/CT analysis and visualization in both procedure planning and guidance stages of bronchoscopy, while the system enables true integration, fusion, and visualization of PET/CT image, bronchoscopic video, and EBUS data sources encountered in the lung-cancer staging work flow. We tested all proposed methods using a ground-truth database established from over 30 lung-cancer patient studies. Finally, we successfully validated our complete multimodal system through a series of lung-cancer patient studies and live guided bronchoscopy procedures.