Comparing Brain Tumor Classification Using Machine Learning

COMPARING BRAIN TUMOR CLASSIFICATION USING MACHINE LEARNING METHODS ENHANCED WITH IMAGE FILTERING (ANISOTROPIC, GAUSSIAN FILTER, NLM)

Abstract

Brain tumor classification is a critical task in medical imaging, aiding in the timely diagnosis and treatment of brain abnormalities. This study evaluates and compares the performance of machine learning models for brain tumor classification, enhanced with three distinct image filtering techniques: Non-Local Means (NLM) filtering, Anisotropic filtering, and Gaussian filtering. Each filtering method is applied during preprocessing to improve image quality by reducing noise and enhancing critical features such as edges and textures. A custom convolutional neural network (CNN) is trained and tested on a publicly available brain tumor dataset, comprising glioma, meningioma, pituitary tumors, and normal brain images. The preprocessing pipeline includes resizing, denoising, and contrast enhancement, tailored for each filter. Experimental results demonstrate that the integration of advanced filtering techniques significantly influences classification accuracy. NLM filtering, with its ability to preserve texture and detail while reducing noise, outperformed both Anisotropic filtering, which excels in edge preservation, and Gaussian filtering, a simpler approach primarily focused on noise reduction. Metrics such as accuracy, precision, recall, and F1-score were used to benchmark the models. The findings suggest that sophisticated filtering methods like NLM and Anisotropic filtering can enhance the reliability of automated brain tumor classification systems. This comparative analysis highlights the role of preprocessing in improving the robustness and effectiveness of machine learning applications in medical imaging.

Keywords:

Brain Tumor Classification, Machine Learning, Image Filtering, Non-Local Means Filter, Anisotropic Filter, Gaussian Filter, Medical Imaging, Convolutional Neural Networks (CNN), Preprocessing, Noise Reduction, Edge Preservation, Texture Enhancement.