Classification of Brain Tumor Using an Optimized Deep Learning Technique to Correlate with Disease State

eng Oriental Scientific Publishing Company Biomedical and Pharmacology Journal 0974-6242 2025-03-31 18 1 983 997 10.13005/bpj/3147 64668 article Classification of Brain Tumor Using an Optimized Deep Learning Technique to Correlate with Disease State Muniraj Gupta 1 Sheetal Bhatia 2 Naveen Sharma 3 Nidhi Verma 4 Saurabh Kumar Sharma 1 Rajkumar Brojen Singh 5 School of Computer and Systems Sciences, Jawaharlal Nehru University, New Delhi, India Department of Biotechnology, Hemvati Nandan Bahuguna Garhwal University, Dehradun, Uttarakhand, India Indian Council of Medical Research, New Delhi, India Department of Microbiology, Ramlal Anand College, University of Delhi, New Delhi, India School of Computational & Integrative Sciences, Jawaharlal Nehru University, New Delhi, India Brain tumor classification is a crucial task in medical image analysis due to the complexity of the neurological system. With the rapid advancements in deep learning techniques, particularly in medical imaging, there is growing potential to enhance the accuracy and efficiency of brain tumor diagnosis using the magnetic resonance imaging (MRI). This paper proposes an optimized and low-computation deep learning model built on the backbone MobileNetv2 convolutional neural network architecture to classify the brain tumors into three categories: glioma, meningioma, and pituitary tumors. The model is trained, validated, and tested using a dataset of T1-weighted contrast-enhanced brain MR images (T1W-CE MRI). Preprocessing steps are incorporated to enhance the classification efficiency. We evaluate the model's performance on both equally and unequally distributed classes of the images and achieve an accuracy of 92.23% and 93.59%, with F1 scores of 92.21% and 93.65%, respectively, for both the distributions. The experimental results demonstrate that the proposed model efficiently classifies the brain tumors using the MR images and achieves superior accuracy to the latest literature methods and state-of-the-art models: "ResNet50, VGG16, NASNetMobile, InceptionResNetV2, and InceptionV3". Thus, the proposed model may help assisting the radiologists in fast and better diagnoses. https://biomedpharmajournal.org/vol18no1/classification-of-brain-tumor-using-an-optimized-deep-learning-technique-to-correlate-with-disease-state/ Brain tumor Classification Convolution Neural Network Deep Learning MR images.