Related Experiment Video
Updated: May 25, 2026
12:08
From Voxels to Knowledge: A Practical Guide to the Segmentation of Complex Electron Microscopy 3D-Data
Published on: August 13, 2014
A graph-theoretic approach for segmentation of PET images
Ulaş Bağci1, Jianhua Yao, Jesus Caban
1Center for Infectious Disease Imaging, Department of Radiology and Imaging Sciences, National Institutes of Health, MD, USA. ulas.bagci@nih.gov
Summary
This study introduces a graph-based method for segmenting positron emission tomography (PET) images, improving accuracy and reliability in disease diagnosis. The new approach outperforms traditional threshold-based methods in segmentation tasks.
Area of Science:
- Medical Imaging
- Radiology
- Image Analysis
Background:
- Accurate segmentation of positron emission tomography (PET) images is crucial for quantifying radio-tracer activity in regions of interest, vital for disease diagnosis and treatment monitoring.
- Standardized uptake values (SUVs) derived from PET scans require precise delineation of functional volumes for reliable analysis.
Purpose of the Study:
- To present a novel graph-based method for robust and accurate segmentation of functional volumes in PET images.
- To evaluate the performance of this method against established threshold-based techniques and assess its reproducibility with clinical data.
Main Methods:
- A graph-based segmentation approach was developed and applied to PET images.
- The method was validated using PET phantoms with ground truth CT simulations.
- Performance was compared against two common threshold-based segmentation methods.
- Intra- and inter-observer variations were assessed using real clinical PET data.
Main Results:
- The graph-based segmentation method demonstrated superior accuracy, robustness, and repeatability compared to threshold-based methods.
- The proposed method showed higher computational efficiency.
- Reproducibility assessments on clinical data confirmed the reliability of the segmentation.
Conclusions:
- The graph-based method offers a significant advancement in PET image segmentation, providing more accurate and reliable results.
- This technique holds promise for improving quantitative analysis in clinical PET imaging.
- The method's superiority suggests potential for wider adoption in diagnostic and research settings.
