GRU Based Autoencoder with Maximum Overlap Discrete Wavelet Transfrmers for Brain Tumor Detection

A new approach based on autoencoder with GRU (gated recurrent unit) and discrete transforms for multi scale features of MRI images. Most of the systems fail to adequately handle sequentially dependencies or effectively leverage multi resolution scale. Therefore, in this paper we are analyzing a spatial and temporal features of brain tumor MRI images by imposing modified wavelet transforms for multi-scale analysis with a sequential modeling approach. Masoud Nickparver's Kaggle is used to train and test this model. In this study, 80% of the data are used for training and 20% of the data is used for testing. The implementation outputs give strong validation by achieving 98.88% classification accuracy for four classes. The F1 score and Precision of our novel framework are 98% and 98.30% respectively.