Bulletin of Electrical Engineering and Informatics Vol. 14, No. 4, August 2025, pp. 2849~2860 ISSN: 2302-9285, DOI: 10.11591/eei.v14i4.9142  2849 Journal homepage: http://beei.org Efficient brain cancer identification using ResNet50 and ResNet50 V2: a comparative study with a primary MRI dataset Md Mizanur Rahman, Israt Jahan, Rana Das, Syada Tasmia Alvi, Chowdhury Abida Anjum Era, Atik Asif Khan Akash, Amir Sohel, Zahura Zaman Department of Computer Science and Engineering, Faculty of Science and Information Technology, Daffodil International University, Dhaka, Bangladesh Article Info ABSTRACT Article history: Received Aug 13, 2024 Revised Jan 28, 2025 Accepted Mar 9, 2025 Primary malignant brain tumors along with central nervous system cause a significant amount of deaths every year, making brain cancer a major worldwide health problem. In South Asian countries, the number of patients suffering from brain cancer is steadily rising. Treatment effectiveness and improved patient outcomes depend on early detection. Using a dataset consisting of 6056 original raw MRI scans, this study evaluates how well convolutional neural networks (CNNs) diagnose brain cancer. We present ResNet50 and ResNet50V2 models assessed for their effectiveness in identifying brain cancers. Transfer learning and fine-tuning were employed to enhance model performance. The models demonstrated strong performance, with 87-99% accuracy rate. ResNet50V2 achieved the highest 99% accuracy. To detect tumor early, this work emphasizes how well the CNN-based machine learning methods help as timely intercession and patient care is necessary. Early prediction with 100% confidence and reliable precision is a critical issue in the modern world. Our goal is to use advanced algorithms to forecast images affected by cancer. Lastly, we will deploy an automated system that will enable us to confidently identify images affected by cancer. Our suggested methodology and its application could significantly impact the field of medical science by combining computer vision and health informatics. Keywords: Brain cancer Cancer predictor Deep learning Image processing MRI images ResNet50 ResNet50V2 This is an open access article under the CC BY-SA license. Corresponding Author: Md Mizanur Rahman Department of Computer Science and Engineering, Faculty of Science and Information Technology Daffodil International University Dhaka, Bangladesh Email: mizanurrahman.cse@diu.edu.bd 1. INTRODUCTION The human brain, which is an essential organ, is responsible for the entire process of control and decision-making. This section needs to be kept safe from injury and disease since it is the controlling center of the nervous system [1]. Every part's final stage is the reason for mortality in cases of different types of brain cancer or tumors. Hence, focusing on prevention or treatment is crucial, particularly in the brain organ. One of the conditions that can immediately endanger people's lives is brain tumors. By 2023, the American Cancer Society predicts that 25,400 cases of cancerous brain or spinal cord tumors will be found. Of these cases, 14,420 will be in men and 10,980 will be in women. These figures would be substantially higher if benign tumors (tumors other than cancer) were included. Nearly 18,760 persons (10,690 men and 8,070 women) lost their lives to malignancies of the brain and spinal cord every year [2]. We gather a healthy dataset on brain tumors, which include meningitis, glioma, and other tumors classified as brain cancer. The