Carleton University
Development of an Integrated System for Automatic Tumor Detection for PET-CT Images
Abstract
dc:description.abstractAccurate early cancer diagnosis helps guide optimal treatment. Hybrid imaging of 18F-lablelled-fluorodeoxyglucose (FDG) using hybrid positron emission tomography (PET) and x-ray computed tomography (CT) is highly sensitive for detecting and characterizing many cancer types with increasing clinical use. Having an effective and reliable tumor detection and segmentation system is a major challenge because of the wide variations in the clinical environments and tumors shapes and sizes. Therefore, an automated clinical system must address three requirements: (1) seamless integration with clinical systems, (2) reliable performance across varying datasets, and (3) handling of extreme conditions outside performance limits. We have developed an automated system composed of three main subsystems: (1) A system to classify CTs axially into anatomical regions, (2) multi-organ segmentation system, and (3) tumor detection/segmentation system. The anatomical classifier reduces the search space for organ segmentation, and the organ segmentation system reduces the search space for tumors. The subsystems manage variations in image representation such as intensity, imaging field of view, and reconstruction parameters. We have developed our own analytical solutions combined with deep learning, achieving robust performance across image types. The systems standardize and inverse-standardize the image space before and after processing. Standardization transforms images into a standard format on which the detection or segmentation system is optimally trained. After getting the results, inverse-standardization transforms the images to the original format. The anatomical system successfully classified body regions with accuracies > 99%. The multi-organ segmentation system successfully segmented lungs, liver, spleen, and kidneys exceeding 90% dice with potential for optimizations. The tumor detection and segmentation system was tested for liver and lungs with detection accuracy of 87% and 80% and overall true-tumor dice of 82% and 91%, respectively. We identified sources of limitations, mainly related to small and false-positive tumors. We proposed three promising approaches to handle false-positives separately. Our system can integrate with clinical environment using digital imaging and communications in medicine (DICOM) import/export functionality. The modular design provides relatively easy and rapid expansion to new applications including other image modalities like magnetic resonance imaging (MRI).
Degree
thesis:*- Name thesis:degree_name
- Doctor of Philosophy (Ph.D.)
- Level thesis:degree_level
- Doctoral
- Discipline thesis:degree_discipline
- Engineering, Electrical and Computer
- Grantor dc:publisher
- Carleton University
- Year dc:date.issued
- 2021
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Salman, Odai S.
Subjects
dc:subject × 2Rights
dc:rights- Statement dc:rights
-
- Copyright © 2021 the author(s). Theses may be used for non-commercial research, educational, or related academic purposes only. Such uses include personal study, distribution to students, research and scholarship. Theses may only be shared by linking to the Carleton University Institutional Repository and no part may be copied without proper attribution to the author; no part may be used for commercial purposes directly or indirectly via a for-profit platform; no adaptation or derivative works are permitted without consent from the copyright owner.
- Language dc:language.iso
- en