Lightweight CNN for Fast and Accurate CT/MRI brain Image Classification

Abstract

Accurate and rapid diagnosis of brain tumors is crucial for improving patient outcomes. However, traditional methods rely on manual examination of MRI and CT images, a process that is time- consuming and labor-intensive and whose accuracy relies heavily on the expertise of specialists, which can lead to variability in results. Technologies such as machine learning and deep learning have emerged as promising alternatives to automate this process. However, current models often require significant computational resources, limiting their use in immediate diagnosis or their generalization to end-user devices. This presents a challenge for real-world deployment. This limitation has created a competitive environment for modifying existing models or developing effective new ones. To address this, our research aims to develop a lightweight convolutional neural network capable of balancing accuracy and efficiency, enabling faster and easier classification of brain images. We used techniques such as image resizing and data augmentation to provide a suitable training environment for our model. We ensured our model's ability to generalize and provide excellent performance with both visible and unseen data, using techniques such as [Dropout] and Cross- validation. Finally, we compared the model's performance with pre-trained models [VGG16] and [ResNet50]. Experimental results showed that our model achieved an accuracy of 98% a sensitivity of 98%.