Deep learning has developed into a very active and practical research tool that is applied in many domains of image processing. Pre-processing is an enhancement of the picture data that reduces unintentional distortions or boosts certain elements of the image that are crucial for subsequent processing. By extracting the supporting information from many multimodality medical images, the image fusion approach is one of the most significant and beneficial medical disease investigation tools. Combining photos from several imaging methods that is magnetic resonance imaging and computed tomography to improve the photo content is known as medical image fusion. The resulting fused images via image fusion will be more illuminating than the individual input images, making them better suited for classification issues. In this paper, to create fused pictures, brain tumor images are used with a discrete cosine transform-based fusion approach. The brain tumor is characterised by unchecked tissue development. If discovered at an early stage, it is easily curable. This study examines how image fusion can enhance the quality of the final images and hence enhance classifier performance.