Related Experiment Videos
Usability of semiautomatic segmentation algorithms for tumor volume determination
A Mahr1, S Levegrün, M L Bahner
1Department of Medical Physics-E0400, Deutsches Krebsforschungszentrum, Heidelberg, Germany.
Investigative Radiology
|February 10, 1999
Summary
Semiautomatic segmentation algorithms show potential for tumor volumetry, with region-growing and isocontour methods being the most promising. Further development is needed for clinical application.
Area of Science:
- Medical imaging analysis
- Computational anatomy
Background:
- Tumor volume is crucial for clinical decisions.
- Current tumor volumetry lacks standardized semiautomatic segmentation methods.
Purpose of the Study:
- To evaluate the clinical usability of semiautomatic segmentation algorithms for tumor volume determination.
- To assess accuracy, contour variability, and time performance of various algorithms.
Main Methods:
- Tested region-growing, volume-growing, isocontour, snakes, hierarchical, and histogram-based algorithms.
- Used an organic phantom simulating liver and metastases for testing.
- Measured real tumor volumes via water displacement as the gold standard.
Main Results:
- Volume variability ranged from 3.9% (isocontour) to 11.5% (hierarchical).
- Segmentation time per slice varied from 32 seconds (volume-growing) to 72 seconds (snakes).
Conclusions:
- Region-growing and isocontour algorithms show potential for tumor volumetry.
- Further algorithm refinement is required for clinical implementation.