Three Dimensional Route Planning for Medical Image Reporting and Endoscopic Guidance
Open Access
- Author:
- Gibbs, Jason David
- Graduate Program:
- Electrical Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- July 07, 2008
- Committee Members:
- William Evan Higgins, Committee Chair/Co-Chair
Zhiwen Liu, Committee Member
Jeffrey Louis Schiano, Committee Member
Robert T Collins, Committee Member - Keywords:
- image processing
image-guided surgery
virtual endoscopy
bronchoscopy
route planning - Abstract:
- Lung cancer is the deadliest form of cancer in the United States. The treatment of lung cancer often begins with the identification of a suspect region of interest (ROI) in multidetector computed tomography (MDCT) chest scans. Such suspicious ROIs may be lymph nodes, masses, infiltrates, or solitary pulmonary nodules. The state-of-the-art process for assessing the ROIs involves off-line procedure planning in the three-dimensional (3D) MDCT image followed by videobronchoscopy. Unfortunately, the current bronchoscopic planning practice is a tedious, error-prone process, requiring significant effort by the physician. These difficulties have prompted physicians to call for automated solutions for the planning and guidance of bronchoscopic procedures. This thesis presents the first complete approach for automatically planning bronchoscopic procedures. In this paradigm, the only interaction required by the physician is the selection of ROI locations. To find safe, appropriate routes to the ROIs, the planning methodologies account for physical, device, and anatomical constraints. The planning calculations are computationally efficient, fit smoothly within the clinical work flow, and produce routes that enable effective bronchoscopy. Because our planning methods rely upon accurate patient-specific airway-tree models derived from the MDCT chest scan to find appropriate bronchoscopic routes through the airway tree, this thesis also presents improvements to the organ model that help enable effective route planning. We present methods to better segment the airway tree from the MDCT image and quantitatively measure the airway cross sections to determine the locations within the airway tree that are accessible by a videobronchoscope of a given diameter. A key improvement, which has allowed physicians to perform peripheral bronchoscopic procedures deeper into the anatomy than ever before, is a new airway-tree surface-definition method. These polygonal airway surfaces are vital in the creation of image-based reports and in the live guidance of bronchoscopy to convey the patient's anatomy to the physician. Results demonstrate the efficacy of the planning systems and airway-tree model improvements. Our automated planning data, when incorporated into a bronchoscopic guidance system, increased the effectiveness of peripheral bronchoscopic procedures in a phantom study. The performance of bronchoscopists of varying levels of experience increased from 43% under the standard practice, to 94% when following the routes of this thesis to ROIs within the phantom. Our planning and airway-tree model data has also been used in live human procedures. In this study, we demonstrate that the entire planning process, including the complete definition of the airway-tree model, can be completed in under 15 minutes, on average. Using the planning and airway-tree model data of this thesis in image-based reports and in live bronchoscopic guidance, physicians have been able to reliably reach peripheral ROIs located deeper within the patient's anatomy than those reached using the best previously-existing manual planning practices.