A Weighted Ensemble Approach for Enhanced Brain Tumor Detection and Classification Using CNN and VGG16*

Detection and classification of brain tumors are essential tasks in medical imaging, focused on enhancing diagnostic accuracy and facilitating early intervention. This study presents a novel weighted ensemble method that integrates the predictive capabilities of a traditional Convolutional Neural Network (CNN) with a pre-trained VGG16 model for effective brain tumor classification. The dataset employed comprises MRI images categorized as 'tumor' or 'no tumor.' Both models were trained on the data, achieving notable results individually, with the CNN attaining an accuracy of 88% and VGG16 reaching 84%. To enhance performance, we utilized a weighted ensemble strategy, allocating a greater weight to the custom CNN due to its superior performance. The weighted ensemble method demonstrated superior classification performance, achieving an accuracy of 92%, precision of 88%, recall of 100%, and an F1-score of 93%. The proposed ensemble surpassed individual models and a stacking ensemble in critical metrics, notably in recall, achieving a perfect score and ensuring the accurate identification of all positive cases. The results demonstrate the efficacy of weighted ensemble methods in improving brain tumor detection, presenting a valuable tool for clinical diagnostics characterized by enhanced reliability and robustness.