Related Experiment Video
Updated: Jan 5, 2026
14:15
Multicolor 3D Printing of Complex Intracranial Tumors in Neurosurgery
Published on: January 11, 2020
7.6K
Postoperative glioma segmentation in CT image using deep feature fusion model guided by multi-sequence MRIs
Fan Tang1,2,3, Shujun Liang1,2, Tao Zhong1,2
1School of Biomedical Engineering, Southern Medical University, No. 1838 Guangzhou Northern Avenue, Baiyun District, Guangzhou, 510515, Guangdong, China.
European Radiology
|October 26, 2019
Summary
A new deep learning model accurately segments postoperative gliomas on CT scans using multi-sequence MRI guidance. This automated method improves segmentation accuracy for radiotherapy planning.
Area of Science:
- Medical imaging
- Artificial intelligence
- Neurosurgery
Background:
- Computed tomography (CT) and magnetic resonance imaging (MRI) are standard for glioma imaging.
- Radiotherapists use multi-modal MRI to delineate CT glioma regions.
- Accurate segmentation is crucial for effective radiotherapy planning.
Purpose of the Study:
- To develop a deep feature fusion model (DFFM) for automated postoperative glioma segmentation in CT images.
- To leverage multi-sequence MRI information to guide CT-based segmentation.
- To enhance the accuracy and efficiency of glioma delineation for radiotherapy.
Main Methods:
- A multi-sequence MRI-guided convolutional neural network (CNN) architecture was developed (DFFM).
- The model simultaneously learns deep features from CT and multi-sequence MRI data.
- A dataset of 59 postoperative glioma CT/MRI scans (grades II-IV) was used for training and validation.
Main Results:
- DFFM achieved a significantly higher Dice coefficient (DSC) of 0.836 compared to U-Net models (0.713 and 0.818).
- Precision values mirrored the DSC improvements, indicating enhanced segmentation accuracy.
- Segmentation performance showed no significant difference across different glioma grades.
Conclusions:
- The DFFM provides accurate automated segmentation of postoperative gliomas on CT images.
- Guidance by multi-sequence MRI significantly improves deep learning-based segmentation.
- This model shows promise for improving and facilitating radiotherapy planning in glioma treatment.
