Abstract – Brain tumors remain a pressing global health concern, with high mortality rates despite significant medical advancements. The brain tumor is a potentially deadly condition, and its categorization creates an enormous challenge for radiologists due to the diverse composition of tumor cells. In recent times, there has been a growing interest in the development of computer-aided diagnostic systems that use magnetic resonance imaging (MRI) to help in the identification of brain tumors. These systems have the potential to provide valuable support in the medical field. In this paper, we have proposed an experimental method for the extraction of brain tumors from 2D Magnetic Resonance Imaging (MRI) scans using a convolutional neural network (CNN). The qualitative study was conducted using a real-time online dataset that included a wide range of tumor sizes, locations, forms, and varying image intensities. The proposed methodology aims to deliver the most effective model capable of distinguishing between normal brain structures and various types of brain tumors, such as glioma, meningioma, and pituitary tumors, using these MRI images. The architecture of the network, the choice of hyperparameters, and the nuances of the optimization algorithms—all these factors play a crucial role in determining the model’s performance. In our work, CNN achieved an accuracy rate of 98.91%, which is quite impressive. In addition, results from experiments demonstrated that the proposed method is efficient and suitable for computer-assisted brain tumor detection.
Cite – Chowdhury, M. J. U., & Kibria, S. (2023, December). Performance Analysis for Convolutional Neural Network Architectures using Brain Tumor Datasets: A Proposed System. In 2023 International Workshop on Artificial Intelligence and Image Processing (IWAIIP) (pp. 195-199). IEEE.