The increasing incidence of brain tumors has underscored the critical need for accurate diagnosis and effective treatment strategies. This study explores advanced methodologies to enhance brain tumor detection and classification. We introduce an innovative convolutional model designed to significantly improve identification accuracy. The performance of several deep learning algorithms, including Inception V3, GoogLeNet, and VGG-19, is meticulously evaluated in the context of brain tumor image classification.A key novelty of this study is the implementation of decision-level fusion, a method not previously explored in this domain. By combining the classification outputs of multiple models, our approach enhances the overall decision-making process, leading to improved accuracy and robustness. This technique allows for the aggregation of diverse perspectives from different models, thereby mitigating individual model weaknesses and capitalizing on their strengths. Our results indicate that these approaches markedly enhance the accuracy (with Inception V3 reaching 98.25%, GoogLeNet 95.36%, and VGG-19 91.24%) and resilience of brain tumor detection and classification systems, laying the foundation for a reliable diagnostic tool.