Percolation Images: Fractal Geometry Features for Brain Tumor Classification

Abstract

Brain tumor detection is crucial for clinical diagnosis and efficient therapy. In this work, we propose a hybrid approach for brain tumor classification based on both fractal geometry features and deep learning. In our proposed framework, we adopt the concept of fractal geometry to generate a “percolation” image with the aim of highlighting important spatial properties in brain images. Then both the original and the percolation images are provided as input to a convolutional neural network to detect the tumor. Extensive experiments, carried out on a well-known benchmark dataset, indicate that using percolation images can help the system perform better.