Brain Tumor MRI Classification Using Weighted Fractional Fourier Transform and K-Nearest Neighbor

Technological advances can accelerate human work in various fields, including the medical field, such as addressing disorders of the human brain. Digital image processing can be applied to problems such as identifying brain tumors, helping medical technicians diagnose them. In this study, the weighted fractional Fourier transform (WFRFT) and K-Nearest Neighbor (KNN) were implemented for the classification of Magnetic Resonance Imaging (MRI) images of brain tumors. There are four types of MRI images used, namely glioma, meningioma, pituitary, and images without brain tumors. The image classification process consists of 2 processes: training and testing. During training, the training images are extracted using WFRFT features and then selected using Principal Component Analysis (PCA) to obtain lower-dimensional image representations. The selected features are then classified using KNN to obtain a KNN classification model. After obtaining the KNN classification model, the next step in the testing process is to evaluate the test images using the model trained during training. The best results in this study identified brain tumors with an average accuracy of 98.125% using WFRFT feature extraction.