Brain-tumor-segmentation
Segmenting brain tumor by using various networks from these papers:
- U-net: Convolutional networks for biomedical image segmentation
- ResUNet-a: A deep learning framework for semantic segmentation of remotely sensed data
- Resunet++: An advanced architecture for medical image segmentation
Project Topic & Goal
The main goal of this project is the brain tumor segmentation.
- As the use of deep learning in the medical field is gradually increasing, high diagnostic performance is being derived.
- Furthermore, in order to utilize medical data, the need to automate data analysis and purification, which was previously handled by experts, is gradually increasing.
- If a high-accuracy and generalizable model can be used, the time that radiologists spend on data analysis can be effectively reduced.
Dataset
We downloaded the dataset from Kaggle.
- Data format: .tif
- Total: 110 dataset (Train/Validation-100, Test-10)
- Ground-truth data: Mask data
Method & Performance
In order to select a model that can improve performance, we compared three models with promising performance in the segmentation task.
Qualitative evaluation (segmented mask visualization)
Quantitative evaluation
