Classification of brain tumors using wavelet transform and Inception v3 convolutional neural network model

Purpose: The performance of the classifier was measured with accuracy, precision, recall and F1-score on the validation set. Theory and Methods: Discrete wavelet transform is a powerful mathematical tool that provides a multi-resolution representation of any signal or image. It is used to extract features from images and reduce image size. Convolutional neural networks, created by modeling the human visual system, are a special type of neural network that uses a grid-like structure to analyze images. Transfer learning is widely used when the sample size of the dataset is small. Results: In this study, a transfer learning-based classification model is presented using wavelet features. Brain tumors were classified using a four-class dataset consisting of glioma, meningioma, pituitary gland and normal brain MRI images, and an accuracy of 99.65% was achieved. Conclusion: As a result, a study was conducted on the use of wavelet transform and convolutional neural networks in the classification of brain MRI images. When wavelet features are used as the input of the Inception V3 network, it is seen that the accuracyperformance increases by 0.34%.