<![CDATA[Brain tumor detection using deep learning has emerged as a crucial approach to improving early diagnosis and treatment planning. This project presents a novel hybrid deep learning model based on the ShuffleNet architecture to enhance the accuracy and efficiency of brain tumor detection from medical images. Traditional machine learning (ML) models rely on hand- crafted features, which are often time-consuming and less effective. Deep learning, on the other hand, automates feature extraction, improving detection accuracy and reliability. The proposed system leverages the ShuffleNet framework, known for its lightweight and high-performance characteristics, making it well-suited for real- time applications. To further enhance the model’s capability, we modified ShuffleNet by removing its last five layers and replacing them with 15 newly designed layers that increase expressiveness and feature extraction capacity. Additionally, we integrated a leaky ReLU activation function in the feature map to mitigate the vanishing gradient problem and improve model generalization. These enhancements result in superior feature representation and higher classification accuracy for brain tumor pathology detection. The dataset used for model training comprises MRI scans labeled with different tumor types. Preprocessing techniques such as normalization, augmentation, and contrast enhancement are applied to ensure robust training. The modified ShuffleNet model demonstrates higher precision, recall, and F1-score compared to traditional CNN-based models, while maintaining computational efficiency. This system can be deployed in real-time clinical settings to assist radiologists in early tumor detection, reducing human error and enhancing diagnostic speed. The integration of deep learning into medical imaging represents a significant step toward automated, accurate, and efficient brain tumor detection, ultimately improving patient outcomes.]]>
