Related Experiment Video
Updated: Sep 3, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.9K
Brain Tumor Segmentation Using Deep Capsule Network and Latent-Dynamic Conditional Random Fields
Mahmoud Elmezain1,2, Amena Mahmoud3, Diana T Mosa3
1Computer Science Department, Faculty of Science, Tanta University, Tanta 31527, Egypt.
Journal of Imaging
|July 25, 2022
Summary
This study presents an automated brain tumor segmentation method using deep capsule networks (CapsNet) and latent-dynamic conditional random fields (LDCRF). The novel approach achieves competitive performance against state-of-the-art methods on benchmark datasets.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computational Neuroscience
Background:
- Brain tumor segmentation is challenging due to significant biological variability.
- Accurate segmentation is crucial for diagnosis, treatment planning, and monitoring.
Purpose of the Study:
- To develop an automated and accurate brain tumor segmentation method.
- To integrate deep capsule networks (CapsNet) with latent-dynamic conditional random fields (LDCRF) for enhanced segmentation.
Main Methods:
- A three-stage process: pre-processing (N4ITK bias correction, intensity normalization), segmentation (CapsNet training on image patches, LDCRF-CapsNet learning from axial slices), and post-processing (thresholding, small region removal).
Main Results:
- The proposed LDCRF-CapsNet method demonstrated superior performance.
- The method achieved competitive results compared to state-of-the-art techniques on BRATS 2015 and BRATS 2021 datasets.
Conclusions:
- The integrated CapsNet and LDCRF approach offers an effective solution for automated brain tumor segmentation.
- This method shows promise for clinical applications requiring precise tumor delineation.
