Leveraging u-net architecture for accurate localization in brain tumor segmentation
This study presents an approach based on deep learning to segment brain tumors in medical imaging accurately. The segmentation of brain tumors plays a crucial role in diagnosing, planning treatments, and monitoring disease progression. However, existing methods have limitations such as time-consumin...
| Main Authors: | , , , |
|---|---|
| Format: | Conference or Workshop Item |
| Language: | English English |
| Published: |
Institute of Electrical and Electronics Engineers Inc.
2023
|
| Subjects: | |
| Online Access: | http://umpir.ump.edu.my/id/eprint/41910/1/Leveraging%20u-net%20architecture%20for%20accurate%20localization.pdf http://umpir.ump.edu.my/id/eprint/41910/2/Leveraging%20u-net%20architecture%20for%20accurate%20localization%20in%20brain%20tumor%20segmentation_ABS.pdf |
| Summary: | This study presents an approach based on deep learning to segment brain tumors in medical imaging accurately. The segmentation of brain tumors plays a crucial role in diagnosing, planning treatments, and monitoring disease progression. However, existing methods have limitations such as time-consuming procedures, inadequate accuracy, and delayed detection. The U-Net model architecture, a widely used convolutional neural network (CNN) for medical image segmentation tasks, was employed to segment brain tumors in CT and MRI scans to overcome these challenges. The performance of the U-Net model was evaluated on datasets consisting of 32, 64, and 128 slices, respectively. The results demonstrated the achievement of the highest percentage of mean Intersection Over Union (IOU), with an impressive 80.89% for brain tumor segmentation. These results outperformed other existing methods. The proposed model exhibits the potential to reduce manual segmentation time and subjectivity while enhancing the accuracy of brain tumor diagnosis, treatment planning, and disease monitoring. This research contributes to the field by addressing the problem of brain tumor detection and showcasing the promising results attained using deep learning techniques. |
|---|