Unsupervised tumour segmentation in PET using local and global intensity-fitting active surface and alpha matting
Ziming Zeng1, Jue Wang, Bernie Tiddeman
1Department of Computer Science, Aberystwyth University, Aberystwyth, UK; Faculty of Information and Control Engineering, Shenyang Jianzhu University, Shenyang, China.
Computers in Biology and Medicine
|September 17, 2013
Summary
This study presents an unsupervised method for segmenting tumors in PET scans, achieving high accuracy by refining boundaries with advanced techniques. The approach improves tumor volume of interest (VOI) calculation, crucial for cancer diagnosis and treatment planning.
Area of Science:
- Medical Imaging
- Radiology
- Computational Biology
Background:
- Accurate tumor segmentation in Positron Emission Tomography (PET) data is critical for effective cancer diagnosis and treatment monitoring.
- Existing segmentation methods often struggle with limited image resolution and partial volume effects inherent in PET scans.
- Unsupervised approaches are desirable to reduce reliance on manual delineation and improve consistency.
Purpose of the Study:
- To develop an unsupervised method for precise tumor segmentation in PET images.
- To address challenges posed by limited resolution and partial volume effects.
- To achieve sub-voxel precision in calculating volumes of interest (VOIs).
Main Methods:
- An improved anisotropic diffusion filter for noise reduction.
- Hierarchical local and global intensity active surface modeling for initial segmentation.
- Alpha matting for fine-tuning segmentation boundaries.
- Validation using real head-and-neck cancer PET images and phantom data.
Main Results:
- The proposed method demonstrates superior segmentation accuracy compared to previous automatic approaches.
- Successful validation on both clinical patient data and objective phantom data.
- Achieved sub-voxel precision in VOI computation, accounting for image limitations.
Conclusions:
- The developed unsupervised approach offers a robust and accurate solution for tumor segmentation in PET imaging.
- The method effectively handles image noise and partial volume effects, leading to improved VOI precision.
- This technique has the potential to enhance cancer diagnosis and treatment planning through more reliable segmentation.