International Research Journal of Education and Technology

This project presents a novel approach for brain image pathology detection using a hybrid deep learning algorithm, specifically employing the ShuffleNet architecture to improve system accuracy. Brain tumors (BTs), which are rapidly proliferating globally, pose a significant health threat, leading to thousands of fatalities annually. Therefore, the accurate detection and classification of brain tumors are crucial for effective treatment. Traditional machine learning (ML) classifiers often rely on hand-crafted features, which can be time-consuming and less effective. In contrast, deep learning (DL) methods excel in automated feature extraction, making them increasingly popular for classification and detection tasks,In this work, we propose a hybrid deep learning model termed DeepTumorNet, designed for the classification of three types of brain tumors: glioma, meningioma, and pituitary tumors. The model is built upon the ShuffleNet architecture, known for its efficiency and lightweight nature, making it suitable for real-time applications. We removed the last five layers of ShuffleNet and replaced them with 15 newly designed layers to enhance the model's expressiveness. Additionally, we implemented a leaky ReLU activation function in the feature map to further improve performance.The proposed model was rigorously tested on a publicly available research dataset, achieving remarkable results with an accuracy of 99.67%, precision of 99.6%, recall of 100%, and an F1-score of 99.66%. The results demonstrate that DeepTumorNet outperforms existing state-of-the-art models, including AlexNet, ResNet50, DarkNet53, ShuffleNet, GoogleNet, SqueezeNet, ResNet101, EfficientNet, and MobileNetV2, confirming its superiority in brain tumor classification from MRI images.

Keywords: deep learning; brain tumor; MRI; transfer learning; convolutional neural network