On the Performance of Deep Transfer Learning Networks for Brain Tumor Detection Using MR Images

A brain tumor need to be identified in its early stage, otherwise it may cause severe condition that cannot cured once is progressed. precise diagnosis of can play an important role start the proper treatment, which eventually reduces survival rate patient. Recently, deep learning based classification method popularly used for detection from 2D Magnetic Resonance (MR) images. In this article, several transfer methods are analyzed using number traditional classifiers detect tumor. The investigation results on a labeled dataset with images both normal- and abnormal brain. For learning, seven such as VGG-16, VGG-19, ResNet50, InceptionResNetV2, InceptionV3, Xception, DenseNet201. Each them followed by five classifiers, Support Vector Machine, Random Forest, Decision Tree, AdaBoost, Gradient Boosting. All combinations feature extractor classifier investigated evaluate relevant performance terms accuracy, precision, recall, F1-score, Cohen’s kappa, AUC, Jaccard, Specificity. Later on, curves all achieved highest accuracies were presented. presented show best model accuracy 99.39% 10-fold cross validation. article expected useful selection suitable detection.