Improving Brain Tumor Diagnosis Using a Combination of Machine Learning and Ensemble Learning

Background and Aims: Accurate diagnosis of brain tumors using MRI images is critical for enhancing diagnostic precision and guiding appropriate treatment strategies. Traditional machine learning algorithms often struggle to achieve high accuracy due to the complex and heterogeneous nature of tumor structures. This study proposes a hybrid machine learning framework that combines ensemble learning strategies to improve the precision and reliability of brain tumor diagnosis. Methods: The dataset comprised ۲۵۵ T۱-weighted MRI images, including ۹۸ slices from healthy brain tissue and ۱۵۵ slices from tumorous brain tissue. Preprocessing steps were employed to reduce noise and enhance image quality, followed by the extraction of geometric features to characterize tumors. Extracted features included area, perimeter, convex area, solidity, equivalent diameters of major and minor axes, and eccentricity. Three machine learning models—Decision Tree, Random Forest, and Multi-Layer Perceptron—were implemented as base classifiers. An ensemble learning model was subsequently developed by aggregating the predictions of these classifiers, leveraging their complementary strengths while mitigating individual weaknesses. Results: Experimental findings demonstrated that the ensemble model outperformed the individual classifiers. Specifically, the Decision Tree, Random Forest, and Multi-Layer Perceptron achieved classification accuracies of ۸۳%, ۸۰%, and ۷۴%, respectively. The ensemble model surpassed these, achieving an accuracy of ۸۵%, thereby underscoring its superior effectiveness in brain tumor diagnosis. Conclusion: The proposed hybrid framework successfully integrates machine learning and ensemble techniques, offering a robust and reliable solution for brain tumor diagnosis using MRI images. This study underscores the value of combining multiple algorithms to enhance diagnostic accuracy. Future research could further refine this framework by incorporating advanced deep learning methodologies and additional feature extraction techniques to improve both performance and clinical utility.

کلیدواژه ها:

BrainTumor ، MRIImages ، MachineLearning ، EnsembleLearning ، Diagnosis