Improving Effectiveness of Different Deep Transfer Learning-Based Models for Detecting Brain Tumors From MR Images

Early classification of brain tumors from magnetic resonance imaging (MRI) plays an important role in the diagnosis such diseases. There are many diagnostic methods used to identify brain. MRI is commonly for tasks because its unmatched image quality. The relevance artificial intelligence (AI) form deep learning (DL) has revolutionized new automated medical diagnosis. This study aimed develop a robust and efficient method based on transfer technique classifying using MRI. In this article, popular architectures utilized tumor system. pre-trained models as Xception, NasNet Large, DenseNet121 InceptionResNetV2 extract features experiment was performed two benchmark datasets that openly accessible web. Images dataset were first cropped, preprocessed, augmented accurate fast training. Deep trained tested three different optimization algorithms (ADAM, SGD, RMSprop). performance evaluated metrics accuracy, sensitivity, precision, specificity F1-score. From experimental results, our proposed CNN model Xception architecture ADAM optimizer better than other models. achieved precision specificity, F1-score values 99.67%, 99.68%, 99.66%, 99.68% MRI-large dataset, 91.94%, 96.55%, 87.50%, 87.88%, 91.80% MRI-small respectively. superior existing literature, indicating it can be quickly accurately classify tumors.