Related Experiment Video
Updated: Dec 30, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.3K
Brain-vascular segmentation for SEEG planning via a 3D fully-convolutional neural network
Summary
This study presents an automated method for segmenting brain blood vessels in medical imaging using a residual Fully Convolutional Neural Network (FCNN). The approach achieves accurate and fast results, aiding in surgical planning and decision-making.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Neurosurgery
Background:
- Three-dimensional visualization of vascular structures is crucial for surgical planning and guidance.
- Accurate segmentation of brain vessels from medical imaging data remains a challenge.
Purpose of the Study:
- To develop an automatic, accurate, and fast method for brain vessel segmentation.
- To utilize a residual Fully Convolutional Neural Network (FCNN) for segmenting Contrast Enhanced Cone Beam Computed Tomography (CE-CBCT) datasets.
Main Methods:
- A residual Fully Convolutional Neural Network (FCNN) with an encoder-decoder architecture was employed.
- A two-stage training strategy was implemented to address dataset imbalance.
- The FCNN was trained using mini-batch gradient descent with the Adam optimizer and binary cross-entropy loss.
Main Results:
- The FCNN achieved a median Dice score of 0.79, Precision of 0.8, and Recall of 0.69.
- Performance was evaluated on 5 CE-CBCT datasets against manual annotations.
Conclusions:
- The proposed residual FCNN method offers an effective solution for automated brain vessel segmentation.
- This technique can enhance preoperative planning, intra-operative guidance, and post-operative decision-making in neurosurgery.
