Detection of Brain Tumor using CNN with PCA and Transfer Learning

Brain tumours demonstrate a wide variety of visuals and possess many features with typical cells, making the removal of regions of tumour from images a difficult endeavour. Moreover, It requires exhaustive time and work to physically process a big amount of data. In order to train the CNN model, this study used transfer learning to create a model for identifying brain tumours from two-dimensional MRI scans. The study employed the Kaggle dataset, which contains tumours with numerous dimensions, sizes, and magnitudes. Unusual regions of tumour have been successfully eradicated from the brain using deep learning, amongst the most sophisticated specialized approaches for tumour categorization and identification. After extracting image features through deep learning, dimension reduction is accomplished using the PCA algorithm. In particular, to extract image features, we first use a deep convolutional neural network. Next, we introduce and apply the PCA algorithm to achieve dimension reduction. Following that, we used the Hyper Parameter Tuning process on CNN, and the accuracy significantly improved.