A Hybrid Deep CNN-SVM Approach for Brain Tumor Classification

Background: Feature extraction process is noteworthy in order to categorize brain tumors. Handcrafted feature consists of profound limitations. Similarly, without appropriate classifier, the promising improved results can’t be obtained. Objective: This paper proposes a hybrid model for classifying tumors more accurately and rapidly preferable choice aggravating tasks. The main objective this research classify through Deep Convolutional Neural Network (DCNN) Support Vector Machine (SVM)-based model. Methods: MRI images are firstly preprocessed improve following steps: resize, effective noise reduction, contrast enhancement. Noise reduction done by anisotropic diffusion filter, enhancement adaptive histogram equalization. Secondly, implementation augmentation enhances data number variety. Thirdly, custom deep CNN constructed meaningful extraction. Finally, superior machine learning classifier SVM integrated classification After that, proposed compared with transfer models: AlexNet, GoogLeNet, VGG16. Results: method uses ‘Figshare’ dataset obtains 96.0% accuracy, 98.0% specificity, 95.71% sensitivity, higher than other models. Also, takes less time others. Conclusion: effectiveness CNN-SVM divulges performance, which manifests that it extracts features automatically overfitting problems improves performance structure, time-consuming. Keywords: Adaptive equalization, Anisotropic CNN, E-health, learning, SVM, Transfer learning.