Related Experiment Video
Updated: Oct 5, 2025
06:48
Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
Published on: January 7, 2019
9.1K
Brain tumor segmentation in MRI images using nonparametric localization and enhancement methods with U-net
Ahmet Ilhan1, Boran Sekeroglu2, Rahib Abiyev2
1Computer Engineering Department, Near East University, 99138, Nicosia, Cyprus, TR, Mersin 10, Turkey. ahmet.ilhan@neu.edu.tr.
International Journal of Computer Assisted Radiology and Surgery
|January 29, 2022
Summary
This study introduces a novel deep learning system using U-net for accurate brain tumor segmentation in MRI scans. The method enhances low-contrast tumors, achieving superior segmentation results for improved diagnosis and treatment planning.
Area of Science:
- Medical Imaging Analysis
- Artificial Intelligence in Medicine
- Neuro-oncology
Background:
- Accurate brain tumor segmentation is crucial for diagnosis, monitoring, and treatment.
- Challenges in segmentation include low-contrast tumors and similarity to healthy tissue.
- Advanced AI methods are being developed to improve segmentation accuracy.
Purpose of the Study:
- To propose an efficient system for complete brain tumor segmentation from MRI images.
- To enhance the visual appearance of indistinct or low-contrast tumors.
- To improve the segmentation abilities of deep learning models.
Main Methods:
- A deep learning architecture, U-net, is employed for segmentation.
- Tumor localization is performed using a histogram-based nonparametric method.
- A novel tumor enhancement method modifies localized regions to improve contrast.
- Enhanced images are then fed into the U-net for segmentation.
Main Results:
- The proposed system was tested on benchmark datasets: BRATS 2012, 2019, and 2020.
- Superior Dice scores were achieved: 0.94 (BRATS 2012 HGG-LGG), 0.85 (BRATS 2019), 0.87 (BRATS 2020).
- The methods demonstrated improved segmentation accuracy for brain tumors.
Conclusions:
- The proposed methods significantly enhance the segmentation capabilities of deep learning models.
- The system provides high-accuracy and low-cost segmentation of brain tumors in MRI.
- These methods have potential applications in various medical segmentation tasks.
