Related Experiment Video
Updated: May 12, 2026
07:57
Positron Emission Tomography-based Dose Painting Radiation Therapy in a Glioblastoma Rat Model using the Small Animal Radiation Research Platform
Published on: March 24, 2022
Combining multiple FDG-PET radiotherapy target segmentation methods to reduce the effect of variable performance of
Ross J McGurk1, James Bowsher, John A Lee
1Medical Physics Graduate Program, Duke University, Durham, North Carolina 27705, USA. ross.mcgurk@duke.edu
Medical Physics
|April 6, 2013
Summary
Combining segmentation methods like majority voting (MJV) and probabilistic estimation (STAPLE) improves accuracy for positron emission tomography (PET) imaging. These robust approaches enhance object segmentation performance across various imaging scenarios.
Area of Science:
- Medical Imaging
- Image Segmentation
- Positron Emission Tomography (PET)
Background:
- Accurate segmentation of high-uptake objects in 18F-fluoro-deoxy-glucose PET images is challenging due to inconsistent performance of individual methods.
- Various segmentation techniques exist, but a universally consistent approach for diverse imaging situations remains elusive.
Purpose of the Study:
- To investigate the effectiveness of combining individual segmentation methods using simple majority voting (MJV) and probabilistic estimation (STAPLE).
- To reduce the impact of inconsistent performance of individual segmentation methods in PET imaging.
Main Methods:
- Utilized a National Electrical Manufacturers Association phantom with spherical and irregularly shaped volumes filled with 18F-fluoro-deoxy-glucose.
- Acquired PET images at varying object-to-background contrasts.
- Applied five individual segmentation methods: 40% thresholding, adaptive thresholding, k-means clustering, seeded region-growing, and a gradient-based method.
- Combined segmentations using MJV and STAPLE, assessing accuracy with Dice Similarity Coefficient (DSC) and Symmetric Mean Absolute Surface Distances (SMASDs) against CT ground truth.
Main Results:
- Both MJV and STAPLE demonstrated robust performance, achieving high median DSC values (approx. 0.88) and low SMASDs (approx. 0.5-0.7 mm).
- STAPLE showed statistically significant improvements for spheres, while MJV performed better for irregular shapes.
- Neither method showed significant DSC differences based on grid size (128x128 vs. 256x256), but SMASDs varied.
Conclusions:
- Combining segmentation methods provides a robust strategy for object segmentation in PET imaging.
- Both MJV and STAPLE enhance accuracy and robustness, offering good performance without requiring training datasets.
- MJV is noted for its simplicity, ease of implementation, and speed.
