Hybrid Ann-Curvelet-Densenet Framework for Brain Tumor Mri Classification and Segmentation

Brain tumor detection and classification from MRI scans is a critical task in medical diagnostics, demanding high accuracy and robustness due to the variability in tumor appearance, size, and location. Traditional manual segmentation is time-consuming and prone to human error. Deep learning has shown promise in automating this process with increased reliability. Despite advances, challenges remain in extracting discriminative features from MRI images that represent both local textures and global structures. Existing deep learning models either lack sufficient feature abstraction or impose high computational costs. This study proposes a hybrid deep learning approach combines Artificial Neural Networks (ANN), Fast Discrete Curvelet Transform (FDCT), and Densely Connected Convolutional Networks (DenseNet) to improve brain tumor classification and segmentation from MRI images. First, open-source MRI datasets with labeled brain tumors were collected. Preprocessing involved noise reduction and contrast enhancement for uniformity. Dimensionality reduction was applied to reduce computational complexity. FDCT was used for feature extraction, capturing rich edge and texture details. ANN was employed to refine features, which were then input into DenseNet for final classification and segmentation. The proposed model was evaluated using performance metrics such as accuracy, precision, recall, Dice coefficient, and F1-score. It outperformed traditional models including VGG16, ResNet50, and U-Net in both classification and segmentation tasks, achieving an accuracy of 96.3% and a Dice score of 94.5%.