Enhanced Brain Tumors Segmentation In 3D Multimodal Images Using U-Net3+ And SegNet

Brain tumors is a serious health concern that require precise and effective diagnostic instruments. In order to segment Brain tumors from MRI images, this study investigates the use of deep learning (DL) approaches, like convolutional neural networks (CNNs) with segmentation models. This work creates a unique model that compares the U-Net3+ and SegNet architecture. The bottleneck layers and skip connections to improve the accuracy of feature extraction and segmentation. A dataset comprising several forms of modalities, such as flair(Fluid-Attenuated Inversion Recovery), t1, t1ce(T1-weighted with contrast enhancement), t2 was used to train and verify the model. To reduce overfitting and improve model robustness, image preprocessing methods like scaling and normalization are used. Metrics including accuracy, loss, and validation performance over several epochs were used to assess the model's performance. The test accuracy of U-Net3+ architecture provides better result than SegNet architecture, which suggests that the model is useful in segmenting between various tumors like Edema, Tumor core and Enhancing tumor. The results of this study highlight the potential of deep learning in the field of medical imaging and open the door to automated diagnostic tools that can help with Brain tumors early diagnosis and treatment planning.