Brain Tumor Diagnosis Using Hybrid Pre-trained CNN-SVM

One of the most crucial and challenging jobs in the world of medical imaging is the detection of brain tumors. Manual classification with human assistance can lead to inaccurate predictions and diagnoses. Further, this is a very difficult task when a substantial volume of data must be supported. Brain tumors are highly variable in appearance and similarities between tumors and normal tissue make extraction of tumor regions from the critical image. Early detection of tumors and classifying them, whether cancerous or non- cancerous are very helpful to the patient. We have proposed a hybrid model using pre-trained CNN (Convolutional Neural Network) named as the EfficientNet B3 model and Support Vector Machine(SVM) classifier in this work boosting detection and finally classification of tumor images. There are eight categories of the EfficientNet model i.e. B0-B7. From eight categories, we have used the EfficientNet B3. Because this model gave more accuracy and requires less computation time. In our model, EfficientNet used as feature extractor and SVM used as the classifier. The accuracy of our proposed hybrid model is 99%. The primary objective of this study is to identify categorize tumors into three types: glioma, malignant, and pituitary after correctly identifying them from non-tumor images.