Related Experiment Video
Updated: Apr 6, 2026
10:25
Brain Infarct Segmentation and Registration on MRI or CT for Lesion-symptom Mapping
Published on: September 25, 2019
49.7K
Segmenting the Brain Surface from CT Images with Artifacts Using Dictionary Learning for Non-rigid MR-CT Registration
Information Processing in Medical Imaging : Proceedings of the ... Conference
|July 30, 2015
Summary
This study introduces a new dictionary learning method for segmenting brain surfaces in post-surgical CT scans of epilepsy patients. The approach accurately extracts the brain surface and improves electrode localization for surgical planning.
Area of Science:
- Medical Imaging
- Computational Neuroscience
- Machine Learning
Background:
- Accurate brain surface segmentation is crucial for epilepsy surgery planning, particularly for registering post-operative CT scans with pre-operative MR images.
- Implanted electrodes in CT scans pose challenges for brain surface extraction due to artifacts and missing skull data.
- Existing registration methods struggle with the complex boundary of the brain surface in post-surgical CT images.
Purpose of the Study:
- To develop and validate a dictionary learning-based method for robust brain surface segmentation in post-surgical CT images of epilepsy patients.
- To improve the accuracy of electrode localization for enhanced surgical guidance.
- To overcome limitations of traditional methods in handling artifacts and incomplete skull data.
Main Methods:
- A dictionary learning approach was employed to model the appearance of both normal brain tissue and artifacts along the cortical surface.
- A surface-based registration method was used to align pre-operative MR and post-operative CT brain surfaces.
- The proposed method was evaluated on clinical data from epilepsy patients.
Main Results:
- The dictionary learning method successfully segmented the brain surface from post-surgical CT images, even with skull defects and electrode artifacts.
- The approach demonstrated superior performance in localizing implanted electrodes compared to intensity-based rigid and non-rigid registration methods.
- Accurate brain surface extraction facilitated improved registration between MR and CT modalities.
Conclusions:
- Dictionary learning offers a powerful tool for segmenting complex brain surfaces in challenging medical imaging scenarios.
- The developed method enhances the precision of electrode localization, directly benefiting surgical planning and patient outcomes in epilepsy treatment.
- This work advances the integration of imaging modalities for neurosurgical applications.
