Related Experiment Video
Updated: Aug 5, 2025
Brain Infarct Segmentation and Registration on MRI or CT for Lesion-symptom Mapping
Published on: September 25, 2019
Brain segmentation in patients with perinatal arterial ischemic stroke.
Riaan Zoetmulder1, Lisanne Baak2, Nadieh Khalili3
1Department of Biomedical Engineering and Physics, Amsterdam University Medical Center, Location University of Amsterdam, Amsterdam, the Netherlands; Department of Radiology and Nuclear Medicine, Amsterdam University Medical Center, Location University of Amsterdam, Amsterdam, the Netherlands; Informatics Institute, University of Amsterdam, Amsterdam, the Netherlands.
This study introduces an automated method using convolutional neural networks (CNNs) to segment brain tissues and ischemic lesions in infants with perinatal arterial ischemic stroke (PAIS). The CNN approach offers a feasible way to evaluate brain development and treatment effectiveness in large patient datasets.
Area of Science:
- Neuroimaging
- Medical image analysis
- Artificial intelligence in medicine
Background:
- Perinatal arterial ischemic stroke (PAIS) significantly impacts neurological outcomes in newborns.
- Manual segmentation of brain tissues and lesions in infant MRI is time-consuming and labor-intensive.
- Accurate quantification is crucial for assessing brain development and treatment efficacy.
Purpose of the Study:
- To develop and validate an automated method for segmenting brain tissues and ischemic lesions in infant MRI scans.
- To utilize convolutional neural networks (CNNs) for efficient and accurate analysis of PAIS-affected brains.
- To provide a tool for objective evaluation of brain development and treatment response.
Main Methods:
- Retrospective analysis of 115 PAIS patients' MRI scans (baseline and 3-month follow-up).
- Manual annotation of nine baseline and 12 follow-up scans for reference segmentation.
- Training two CNNs for automatic segmentation of brain tissues and ischemic lesions.
- Quantitative evaluation using Dice coefficient (DC) and mean surface distance (MSD).
- Qualitative assessment of segmentation accuracy and impact of scan quality by experts.
Main Results:
- High Dice coefficients (0.78-0.95) and low mean surface distances (0.10-1.08 mm) achieved for brain tissue segmentation.
- Acceptable Dice coefficients (0.72-0.86) and mean surface distances (1.23-2.18 mm) for ischemic lesions.
- Excellent qualitative agreement for brain tissue segmentation (>85% excellent for baseline, >91% for follow-up) in artifact-free scans.
- Segmentation accuracy for ischemic lesions was rated excellent in 61% of baseline scans.
Conclusions:
- Automated segmentation of brain tissue and ischemic lesions in PAIS patients' MRI scans is feasible.
- The developed CNN method shows potential for evaluating brain development and treatment efficacy.
- This automated approach can facilitate large-scale studies and clinical applications in pediatric stroke.
