Related Experiment Video
Updated: Aug 16, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.9K
Clinical Decision Support Framework for Segmentation and Classification of Brain Tumor MRIs Using a U-Net and DCNN
Nagwan Abdel Samee1, Tahir Ahmad2, Noha F Mahmoud3
1Department of Information Technology, College of Computer and Information Sciences, Princess Nourah bint Abdulrahman University, P.O. Box 84428, Riyadh 11671, Saudi Arabia.
Healthcare (Basel, Switzerland)
|December 23, 2022
Summary
This study introduces a simplified deep convolutional neural network (DCNN) for brain tumor (BT) classification and U-Net for real-time segmentation. The developed computer-aided diagnosis (CAD) system achieves high accuracy in classifying brain tumors from MRI scans.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Brain tumors (BTs) are a rare but deadly cancer.
- Computer-aided diagnosis (CAD) systems for BT classification in MRI are crucial but still developing.
- Accurate and efficient segmentation and classification are key challenges.
Purpose of the Study:
- To develop a lightweight U-Net for real-time brain tumor segmentation.
- To present a simplified deep convolutional neural network (DCNN) for automatic feature extraction and classification of segmented brain tumors.
- To improve the accuracy of brain tumor classification in medical imaging.
Main Methods:
- Implemented a lightweight U-Net deep learning network for precise, real-time segmentation.
- Designed a simplified DCNN architecture with five convolutional layers, ReLU activation, normalization, and max-pooling for classification.
- Validated the framework on the multimodal brain tumor segmentation (BRATS 2015) dataset.
Main Results:
- Achieved a Dice similarity coefficient (DSC) of 88.8%, sensitivity of 89.4%, and classification accuracy of 88.6% for high-grade gliomas on BRATS 2015.
- The segmentation performance is comparable to state-of-the-art methods.
- Improved brain tumor image classification accuracy from 88% to 88.6% compared to previous studies.
Conclusions:
- The proposed CAD framework demonstrates competitive performance in brain tumor segmentation.
- The simplified DCNN architecture offers enhanced accuracy for brain tumor classification.
- This research contributes to advancing automated diagnosis systems for brain tumors using MRI.
