The treatment of brain tumors, utilizing conventional methods such as surgery, radiotherapy, and chemotherapy, is limited in terms of accuracy and effectiveness. Furthermore, there exists a possibility of missing the diagnosis for small lesions and certain benign tumors with comparable density to normal tissue. To improve the precision and efficiency of brain tumor diagnosis, recent developments in artificial intelligence have been explored, including the use of Convolutional Neural Networks (CNNs). This research investigates the potential of a four-class CNN-based deep learning algorithm for the diagnosis of brain tumors. A dataset of MRI images, including various forms of brain tumors, underwent preprocessing and cleansing, and was subsequently classified into four categories. The CNN model trained to identify and diagnose MRI images achieved an 85.4% accuracy on the validation set. This study underscores the potential of CNNs to enhance the detection and precision of brain tumors, in addition to improving the consistency and dependability of diagnosis, thereby providing new leads for the discovery of novel therapies and medications. However, the study recognizes that limitations and areas of improvement exist in terms of dataset size, model architecture, and evaluation metrics.