Related Experiment Video
Updated: Mar 12, 2026
04:25
Author Spotlight: Bridging Gaps in Anatomy and Establishing a Foundation for Algorithmic Studies
Published on: December 15, 2023
4.0K
Intra-rater variability in low-grade glioma segmentation
Hans Kristian Bø1,2, Ole Solheim3,4,5, Asgeir Store Jakola3,6,7
1Department of Radiology and Nuclear Medicine, St. Olavs University Hospital, P.O. Box 3250, Sluppen, 7006, Trondheim, Norway. hans.kr.b@gmail.com.
Journal of Neuro-Oncology
|November 13, 2016
Summary
Assessing low-grade glioma (LGG) size and growth via 3D segmentation shows considerable intra-rater variability. This variability decreases with experience and clearer tumor borders, suggesting a need for standardized segmentation criteria.
Area of Science:
- Radiology
- Neuro-oncology
- Medical imaging analysis
Background:
- Accurate assessment of tumor size and growth is crucial for low-grade gliomas (LGGs) prognostication and treatment monitoring.
- The reliability of 3D segmentation for LGGs, which often lack contrast enhancement and have diffuse borders, is not well-established.
Purpose of the Study:
- To investigate the intra-observer variability in segmenting LGGs using 3D FLAIR images.
- To evaluate the impact of experience and tumor characteristics on segmentation reliability.
Main Methods:
- Three independent segmentations of 23 pre-operative LGG 3D FLAIR scans were performed using 3D Slicer by a radiologist without prior experience.
- Intra-rater variability was quantified using the Jaccard coefficient (J2, J3) and Hausdorff Distance (HD).
- Tumor volumes, differences, and relationships with tumor size, border definition, histology, and ADC were analyzed.
Main Results:
- Intra-rater variability was considerable but improved with successive segmentations (J2 improved from 0.87 to 0.90).
- Larger tumors showed a trend towards smaller relative volume differences, and tumors with well-defined borders had significantly less variability (J2, HD).
- No significant relationship was found between variability and histological subtypes or Apparent Diffusion Coefficients (ADC).
Conclusions:
- Significant intra-rater variability exists in serial 3D segmentation of LGGs, particularly for inexperienced users.
- Segmentation reliability improves with experience and is higher for tumors with more conspicuous borders.
- Standardized criteria for defining tumor borders and progression in 3D volumetric segmentation are necessary for reliable LGG assessment.
