Related Experiment Video
Updated: Nov 7, 2025
05:33
Three-Dimensional Shape Modeling and Analysis of Brain Structures
Published on: November 14, 2019
7.3K
Development of a Convolutional Neural Network Based Skull Segmentation in MRI Using Standard Tesselation Language
Rodrigo Dalvit Carvalho da Silva1,2, Thomas Richard Jenkyn1,2,3,4,5, Victor Alexander Carranza1,6
1Craniofacial Injury and Concussion Research Laboratory, Western University, London, ON N6A 3K7, Canada.
Journal of Personalized Medicine
|April 30, 2021
Summary
This study developed a 3D convolutional neural network (CNN) for precise skull segmentation in magnetic resonance imaging (MRI). The CNN achieved improved accuracy after removing the brain region, demonstrating its potential for medical imaging analysis.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Neuroscience
Background:
- Accurate segmentation of anatomical structures is essential in medical imaging analysis.
- 3D Convolutional Neural Networks (CNNs) show promise for automated segmentation tasks.
- Skull segmentation in Magnetic Resonance Imaging (MRI) is challenging but crucial for various applications.
Purpose of the Study:
- To develop and train a 3D CNN for skull segmentation in MRI.
- To evaluate the performance of the CNN using gold standard labels derived from CT scans and Standard Tessellation Language (STL) models.
- To investigate the impact of removing the brain region on segmentation accuracy.
Main Methods:
- Creation of 58 gold standard volumetric labels from CT scans in STL models.
- Conversion of STL models into matrices and overlaying them onto corresponding MR images.
- Training a 3D CNN with 58 MR images and their corresponding gold standard labels.
- Further training with a dataset excluding the brain region, segmented using a 3D CNN and manual corrections.
Main Results:
- The initial 3D CNN achieved a mean ± standard deviation (SD) Dice Similarity Coefficient (DSC) of 0.7300 ± 0.04 for skull segmentation.
- After excluding the brain, the CNN achieved a significantly improved mean ± SD DSC of 0.7826 ± 0.03.
- The study demonstrates the feasibility of using CNNs and STL models for skull segmentation.
Conclusions:
- A 3D CNN framework was successfully developed for skull segmentation in MRI.
- Excluding the brain region enhances the precision of 3D CNN-based skull segmentation.
- This approach offers a valuable tool for medical imaging analysis and research.
