| Autor: |
Shah, Deep, Barve, Amit, Vala, Brijesh, Gandhi, Jay |
| Předmět: |
|
| Zdroj: |
AIP Conference Proceedings; 2024, Vol. 3107 Issue 1, p1-9, 9p |
| Abstrakt: |
Deep learning has produced major advancements in computer vision, and its application for analyzing medical data has expanded considerably. A brain tumor is a disorder in which the brain's surrounding brain cells grow inappropriately and quickly spread to any region of the body. By identifying tumor patterns in brain images and fragmenting the features of brain tumors in those images, brain tumor can be identified. Multi-modal data provides more feature specific weights of same patient. Throughout the experiment we have taken BraTS2018 data set with 285 subjects for the training of the network and every subject contain 4 modalities of brain MRI like T1, T2, T1ce, FLAIR. In this work we propose a network, where a combined modality is generated with the fusion of four different modality features to train the network. It takes 128*128*128 image input size. Segmentation is done with the help of 3D u-net neural network architecture. This approach out-performs different networks such as CNN and other U-Net networks. At the end of the experimentation our proposed approach is found superior than different State of the art approaches. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
