Related Experiment Video
Updated: Dec 14, 2025
12:30
A Comprehensive Protocol for Manual Segmentation of the Medial Temporal Lobe Structures
Published on: July 2, 2014
20.7K
Accuracy and practical aspects of semi- and fully automatic segmentation methods for resected brain areas
Karin Gau1, Charlotte S M Schmidt2,3, Horst Urbach4
1Epilepsy Center, Medical Center - University of Freiburg, Faculty of Medicine, University of Freiburg, Breisacher Str. 64, 79106, Freiburg im Breisgau, Germany. karin.gau@gmail.com.
Neuroradiology
|July 22, 2020
Summary
Semi-automatic segmentation of resected brain areas is more accurate than fully automatic methods for neurological research. This approach offers a cost-effective alternative to manual segmentation, especially for large patient cohorts.
Area of Science:
- Neuroimaging and Medical Image Analysis
- Computational Neuroscience
- Neurosurgery
Background:
- Precise segmentation of brain lesions is crucial for neurological research, aiding in the assessment of postoperative tissue and lesion-symptom mapping.
- Estimating resection volumes is vital for evaluating residual tissue after surgeries, particularly for gliomas and epilepsy.
- Accurate delineation of surgical lesions is fundamental for behavioral lesion-symptom mapping in epilepsy studies.
Purpose of the Study:
- To evaluate the accuracy and cost-efficiency of semi- and fully automatic segmentation methods for resected brain areas.
- To compare ITK-SNAP (semi-automatic) and lesion_GNB (fully automatic) segmentation against manual segmentation as a reference.
- To determine the optimal approach for segmenting resected brain areas in terms of precision and resource utilization.
Main Methods:
- Comparison of semi-automatic (ITK-SNAP) and fully automatic (lesion_GNB) segmentation methods against manual segmentation.
- Accuracy assessment using Dice Similarity Coefficient (DSC) and Hausdorff Distance (HD).
- Evaluation of processing times and inferred costs for each segmentation method using T1w MRI data from 27 epilepsy patients.
Main Results:
- Semi-automatic segmentation demonstrated superior accuracy (median DSC 0.78, median maHD 0.44) compared to fully automatic (median DSC 0.58, median maHD 1.32) (p < 0.001).
- No significant difference was found in median percent volume difference between the two automated approaches (p > 0.05).
- Manual segmentation required substantial time (30.41 min/subject), significantly more than semi-automatic (3.27 min/subject) or fully automatic methods (near-zero labor/cost).
Conclusions:
- Semi-automatic segmentation provides the most accurate results for resected brain areas, balancing precision with moderate human input.
- This method is a viable and cost-efficient alternative to manual segmentation, particularly beneficial for studies involving large patient cohorts.
- The findings support the use of semi-automatic tools for improved efficiency and accuracy in neurosurgical and neurological research.
