Assessing Variability in Brain Tumor Segmentation to Improve Volumetric Accuracy and Characterization of Change
Edgar A Rios Piedra1, Ricky K Taira1, Suzie El-Saden2
1Department of Radiological Sciences at the University of California, Los Angeles, CA. Department of Bioengineering at the University of California, Los Angeles, CA. Medical Imaging Informatics (MII) at the University of California, Los Angeles, CA.
... IEEE-EMBS International Conference on Biomedical and Health Informatics. IEEE-EMBS International Conference on Biomedical and Health Informatics
|July 4, 2017
Summary
Variability in brain tumor segmentation is inherent. Equivalence tests can determine if tumor volume changes significantly over time, improving clinical decision-making for magnetic resonance imaging (MRI) analysis.
Area of Science:
- Medical imaging analysis
- Neurosurgery
- Radiology
Background:
- Volumetric assessment of brain tumors using magnetic resonance imaging (MRI) is crucial for disease progression monitoring.
- Existing segmentation algorithms exhibit variability, potentially leading to misinterpretations of tumor volume changes.
Purpose of the Study:
- To systematically characterize variability in brain tumor boundaries from different segmentation approaches.
- To introduce equivalence testing for determining statistically significant changes in tumor volume over time.
Main Methods:
- Segmentation of 32 MRI studies from 8 patients using four distinct methods: statistical classifier, region-based, edge-based, and knowledge-based.
- Utilized equivalence tests to assess tumor volume changes between sequential time points.
- Calculated average Dice coefficient (0.754) to quantify agreement with a reference standard.
Main Results:
- Demonstrated that variability from different segmentation methods can be leveraged to identify significant tumor volume changes.
- Achieved an average Dice coefficient of 0.754 (95% CI 0.701-0.808) across multiple segmentation approaches compared to a reference standard.
- Highlighted the necessity of considering segmentation variability in the interpretation of tumor volume dynamics.
Conclusions:
- Segmentation variability is an intrinsic aspect of brain tumor analysis.
- Equivalence testing provides a robust method for evaluating true tumor volume changes amidst segmentation variability.
- Incorporating variability assessment enhances the reliability of clinical decision-making in neuro-oncology.
