Related Experiment Video
Updated: Jun 22, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.7K
Detection and Segmentation of Glioma Tumors Utilizing a UNet Convolutional Neural Network Approach with
M Tamilarasi1, S Kumarganesh2, K Martin Sagayam3
1Department of Electronics and Communication Engineering, Sasurie College of Engineering, Tirupur, India.
Summary
This study introduces a novel UNet convolutional neural network (CNN) approach for accurate glioma brain tumor detection. The method enhances tumor segmentation using shearlet transforms, achieving high classification rates on public datasets.
Area of Science:
- Medical Imaging Analysis
- Artificial Intelligence in Oncology
- Neuro-oncology
Background:
- Accurate identification of glioma tumor regions in brain images is crucial for effective treatment and patient outcomes.
- Existing methods may face challenges in precise tumor delineation, impacting clinical decision-making.
- Deep learning architectures like UNet-CNN show promise in medical image segmentation tasks.
Purpose of the Study:
- To develop and validate a robust methodology for prompt and precise glioma brain tumor detection and segmentation.
- To enhance tumor identification by integrating a non-subsampled shearlet transform with the UNet-CNN architecture.
- To evaluate the proposed system's performance on publicly available, challenging glioma datasets.
Main Methods:
- A novel methodology combining a transformation module, feature extraction module, and tumor segmentation module was proposed.
- Brain MRI images were decomposed into low- and high-frequency subbands using a non-subsampled shearlet transform.
- Shearlet features were extracted and classified using the UNet-CNN architecture for tumor region identification.
Main Results:
- The proposed system achieved high mean classification rates: 99.1% on BRATS 2019 and 97.8% on TCGA datasets.
- Excellent performance metrics were observed, including sensitivity (up to 98.2%), specificity (up to 98.7%), and accuracy (up to 98.9%).
- The system demonstrated superior intersection over union (IoU) and disc similarity coefficient (DSC) on both datasets.
Conclusions:
- The integration of non-subsampled shearlet transform with UNet-CNN significantly improves glioma brain tumor detection accuracy.
- The proposed methodology offers a highly effective and efficient approach for precise tumor segmentation in clinical practice.
- Comparative analysis confirms the superiority of this approach over existing state-of-the-art methods for glioma detection.
