Related Experiment Video
Updated: Mar 23, 2026
06:48
Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
Published on: January 7, 2019
9.6K
Automatic Segmentation of MR Brain Images With a Convolutional Neural Network
IEEE Transactions on Medical Imaging
|April 6, 2016
Summary
This study introduces a novel convolutional neural network for automatic brain MR image segmentation. The method achieves accurate and robust segmentation across diverse age groups and imaging protocols.
Area of Science:
- Medical Imaging
- Neuroscience
- Artificial Intelligence
Background:
- Accurate segmentation of brain MR images is crucial for quantitative analysis in large-scale studies.
- Existing methods may struggle with variations in age and acquisition protocols.
Purpose of the Study:
- To develop an automatic segmentation method for MR brain images using a convolutional neural network.
- To ensure accurate segmentation details and spatial consistency across different datasets.
Main Methods:
- A convolutional neural network (CNN) employing multi-scale patch and kernel sizes for voxel-wise classification.
- The method learns features directly from training data, requiring only a single anatomical MR image.
Main Results:
- The method was tested on five diverse datasets, including preterm infants and adults of various ages.
- Average Dice coefficients ranged from 0.82 to 0.91, demonstrating high accuracy.
- The segmentation showed robustness to differences in age and acquisition protocols.
Conclusions:
- The developed CNN-based method provides accurate and robust automatic segmentation of MR brain images.
- This technique is suitable for quantitative analysis across a wide range of ages and imaging parameters.
