Deep Learning-Based Classification of Brain Tumors Using MR Images
Keywords:Convolutional Neural Network (CNN), Magnetic resonance imaging (MRI), Brain tumor, Deep learning, Fast Fourier Transform (FTT), Tamura feature extraction.
In the past several years, academics have focused a lot of attention on Deep Learning (DL), the latest model and most popular trend in the machine learning area. DL has been widely utilized in multiple applications as a strong machine-learning method for handling different complicated issues that call for exceptionally high accuracy and precision, notably in the medical sector. Brain tumors are among the most common and serious malignant tumor illnesses overall, and if they are explored at a more advanced stage, the patients may have a very short life. Thus, grading a brain tumor is an extremely important step to take after finding the tumor for the purpose of developing a successful treatment strategy. In this study, we proposed the Convolutional Neural Network (CNN) model, which is among the most popular structures for deep learning, for classifying the data from an MRI scan that displays brain tumors. The suggested system undergoes a variety of stages, a preprocessing stage that includes (grayscale image transformation, image blurring, histogram equalization, image resize), and feature extraction which includes (Fast Fourier Transform (FFT), K-means Vector Quantization (VQ), and Tamura) and classification stage which accomplish by using the proposed CNN. The experiential results were applied to the Br35H dataset which includes 7023 Magnetic resonance imaging (MRIs) of the human brain. The suggested BTCNN model which consists of 25 layers and is a potent tool that performs overall with 100% accuracy. The proposed system increases the classification accuracy in a very short time.