<![CDATA[Convolutional Neural Networks (CNNs) have become the backbone of various computer vision applications, including medical diagnosis and disease detection. In the context of brain tumor detection, CNNs have demonstrated their capacity to interpret the subtle complexities present in brain images, distinguish between various tumor categories, and provide essential information that can inform clinical decision-making and personalized treatment planning.This study aims to develop a lightweight Convolutional Neural Network (CNN) architecture capable of multiclass brain tumor detection based on RGB images, with a focus on computational efficiency and detection performance. The proposed CNN model adopts a shallow-to-mid depth approach to reduce the number of parameters without sacrificing accuracy. Data augmentation techniques are applied to increase the variability of training images and reduce overfitting, while batch normalization and dropout are used to improve model stability and generalization. The model is trained on an RGB brain tumor image dataset consisting of three tumor classes (glioma, meningioma, and pituitary) and evaluated using accuracy, training time, and the number of parameters to assess computational efficiency. Experimental results show that the developed CNN model achieves an accuracy of over 97% on training and validation data, with efficient training time and a controlled parameter count of approximately 21 million. The main advantage of this model is its computational efficiency, which enables implementation on hardware with limited resources, making it suitable for automated tumor detection systems based on medical imaging. The gap or novelty of this research lies in the development of a lightweight CNN model that is not only resource-efficient but also capable of delivering high-accuracy results in multiclass brain tumor classification tasks using RGB images, while minimizing parameter usage and training time.]]>
