Brain Tumour Prediction Using MRI Images with MobileNet and Transfer Learning: A Deep Learning Approach

This is the authors' abstract. We don't add key points for this paper.

Brain tumour detection and classification are crucial for early diagnosis, improving patient survival rates. Magnetic Resonance Imaging (MRI) offers non-invasive, high-resolution brain images, but manual interpretation is time-consuming and error prone. This research proposes a deep learning approach for brain tumour prediction using MRI images, leveraging the MobileNet architecture with transfer learning, incorporating Flatten, Dense, and Dropout layers to enhance model generalization. MobileNet, a lightweight Convolutional Neural Network (CNN), efficiently processes large-scale MRI datasets. The pre-trained MobileNet is integrated with a Flatten layer to reduce feature map dimensions, followed by Dense layers for learning complex patterns, and a Dropout layer to prevent overfitting. Transfer learning allows leveraging pre-trained weights, reducing the need for large datasets. The model was finetuned on an MRI brain tumour dataset, achieving an accuracy of 0.9845 with a loss of 0.0015, outperforming traditional deep learning methods in tumour classification. This architecture is computationally efficient, making it suitable for real-time clinical applications, providing medical professionals with a reliable, automated tool for brain tumour diagnosis.