Hybridized soccer league-grey wolf Optimization algorithm and siamese network synergism based multimodal fusion scheme for prognosis of brain diseases

• Published in: Multimedia tools and applications Vol. 84; no. 16; pp. 17161 - 17188
• Main Authors: Mukherjee, Suranjana, Banerjee, Sriparna, Chaudhuri, Sheli Sinha
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.05.2025; Springer Nature B.V
• Subjects: Algorithms; Artificial neural networks; Computer Communication Networks; Computer Science; Computer vision; Data Structures and Information Theory; Decomposition; Energy gradient; Entropy; Image quality; Lower bounds; Medical imaging; Multimedia Information Systems; Optimization; Parameters; Principal components analysis; Scaling factors; Soccer; Special Purpose and Application-Based Systems; Track 2: Medical Applications of Multimedia; Keywords is required. Please provide
• ISSN: 1573-7721; 1380-7501; 1573-7721
• DOI: 10.1007/s11042-024-19727-2

This study proposes a novel multi-stage multimodal image fusion approach for consolidating information from images into single scene by leveraging the efficiencies of newly designed hybridized Soccer League (SOL) -G rey Wolf Optimization (GWO) algorithm in association with normalized weight maps generated using dual-branch Siamese Convolutional Neural Network (DSCNN) . The first stage of the designed approach deals with two-scale decomposition of input image (I) into sublayers– Structural Layer (SL) and Residual Layer (RL) containing the gross structural components and edge information of I respectively. The SL is generated through Principal Component Analysis (PCA)-based weighted-averaging of Regional Entropy (REN) and Regional Intensity (RIN). The RL is obtained by subtracting SL from I . In the next step, DSCNN is employed to estimate weight maps for effectively combining pixel level activities of SLs and RLs of multimodal images. The contrast of these estimated weight maps is enhanced using aniso-directional local energy gradient parameter termed as Special Contrast Enhancement Operator. The novelty of this work lies in designing new fusion rule and estimating unique optimum scaling factor for each sublayer by the newly developed SOL-GWO algorithm where upper-lower bound and fitness is characterized by adaptive parameters like Entropy (EN), Standard Deviation (SD), Spatial Frequency (SF), Average Gradient (AG) of decomposed sublayers to increase the method’s robustness The subjective and objective analyses’ conducted using benchmark database and evaluation parameters ( EN, SD, Edge Intensity (EI), SF, Normalized Mutual Information (NMI) and Tone Mapped Image Quality Index (TMQI)) establish the efficiency of the proposed scheme over recent methods.