Fractional Labelmap Representation of Anatomical Structures

Loading...
Thumbnail Image

Authors

Sunderland, Kyle

Date

Type

thesis

Language

eng

Keyword

Fractional Labelmap , Structure Representation , Radiation Therapy , Dose Volume Histogram , Voxelization , Segmentation

Research Projects

Organizational Units

Journal Issue

Alternative Title

Abstract

INTRODUCTION: In medical imaging software, structures are often represented using image volumes known as labelmaps. Conversion of structures to labelmap representations results in a loss of information which impacts medical imaging algorithms, including those in radiation therapy (RT) treatment planning. RT treatment planning systems are used to optimize radiation delivery to structures, which are represented as planar contours. Errors from conversion affect metrics such as dose volume histograms (DVHs) that are the primary metric for RT plan optimization. The goal of this thesis is to develop fractional labelmaps as a structure representation that preserves more structural information and reduces conversion errors. METHODS: The effect of voxelization on treatment planning metrics was tested by comparing DVHs calculated using varying voxel sizes. An algorithm was implemented that triangulates planar contours to closed surface mesh and handles features such as branching, keyhole contours, and end-capping. Closed surfaces were converted into fractional labelmap representations, where each voxel represents occupancy between 0% and 100%. Existing segmentation methods were modified to allow fractional labelmaps to be edited directly. RESULTS: Algorithms were implemented in the open-source medical imaging platform 3D Slicer, and the SlicerRT toolkit. Voxel size was found to affect DVH accuracy for structures with small features or in regions with a high dose gradient. The planar contour to closed surface triangulation algorithm produced qualitatively good surfaces. Fractional labelmaps from closed surfaces were found to be up to 19.1% better at representing structure volume than binary labelmaps, with an average improvement of 6.8%. DVHs calculated from fractional labelmaps were found to be more accurate than DVHs calculated using binary labelmaps. Fractional segmentation methods were found to create good quality segmentations. CONCLUSION: Labelmap voxel size was found to be a contributing factor for DVH accuracy. Accurate conversion algorithms were implemented for planar contour to closed surface and closed surface to fractional labelmap conversions. When used for structure representation and DVH calculation, fractional labelmaps were more accurate than binary labelmaps at the same resolution. Fractional labelmaps were found to be an effective tool for structure representation in radiotherapy that could be expanded to other use cases.

Description

Citation

Publisher

License

Queen's University's Thesis/Dissertation Non-Exclusive License for Deposit to QSpace and Library and Archives Canada
ProQuest PhD and Master's Theses International Dissemination Agreement
Intellectual Property Guidelines at Queen's University
Copying and Preserving Your Thesis
This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.

Journal

Volume

Issue

PubMed ID

External DOI

ISSN

EISSN