Combinatorial Multimodal Fusion to Enhance Brain Tumor Image Using Discrete Wavelet Transform

Multimodal brain imaging combines structural and functional medical images to improve diagnostic accuracy in neurological analysis. However, the fusion of heterogeneous modalities such as MRI, CT, and PET remains challenging due to variations in contrast, spatial resolution, and noise. This paper proposes a novel combinatorial multimodal fusion framework based on the Discrete Wavelet Transform (DWT) that integrates complementary information from the combination of multiple brain imaging modalities. The method employs combinatorial analysis of modalities in different ways, multi-resolution wavelet decomposition, and adaptive fusion rules to maximize structural preservation and functional enhancement. The wavelet decomposition of the dataset has been done at four levels with low and high activity regions.
Quantitative evaluation using Entropy (EN), Mutual Information (MI), Structural Similarity Index (SSIM), Peak Signal-to-Noise Ratio (PSNR), Fusion Factor (FF), and Edge Preservation Index (EPI) demonstrates that the proposed approach outperforms conventional PCA, Laplacian pyramid, Discrete Cosine Transform (DCT), and basic DWT methods. The resulting fused images exhibit superior clarity, contrast, and diagnostic utility. This proposed method gives 98.34% accuracy for the fusion using PSNR. The experiment is tested over the Daubechies (db4) wavelet approach. The quantitative and graphical analysis indicates that the Discrete Wavelet Transform (DWT) significantly improves the quality of fused images among other fusion methods.