MRI images based brain tumor detection using hybrid machine learning and fuzzy technique

Abstract

The detection and classification of brain tumors are essential for improving treatment outcomes and patient survival, making automated diagnostic tools essential. The goal of this project is to create a strong and very accurate machine-learning model that can use MRI data to find and classify brain tumors. The methodology begins with the capture of a well-structured and carefully annotated Kaggle dataset, followed by necessary preprocessing steps like noise reduction, normalization, and contrast enhancement to improve image quality and ensure consistent data distribution. Exploratory Data Analysis (EDA) is performed to identify class discrepancies, manage outliers, and aid with feature engineering choices. We employ traditional machine learning classifiers like Support Vector Machine (SVM), Random Forest, K-Nearest Neighbour (KNN), and XGBoost instead of deep learning classification layers because they are cheaper to compute and work well with small datasets. This research presents an innovative enhancement by integrating Fuzzy C-Means (FCM) Ensemble Learning, which delineates the tumor region by clustering pixels with similar intensity levels, hence promoting superior feature differentiation and reducing classification errors. The experimental findings show a significant improvement in performance, with the proposed FCM Ensemble model achieving an accuracy of 99.58%, which is better than SVM (98.42%), Random Forest (97.25%), and XGBoost (96.88%). It also beats other models that have been released, such the hybrid AlexNet + SVM, which achieved an accuracy of 95.15%. The positive results confirm that the proposed methodology enhances diagnostic reliability, aids in precise tumor localization and classification, and can be effectively employed as a clinical decision support tool to assist radiologists in the prompt and accurate assessment of brain tumors.