Innovative Application of Multimodal Medical Imaging in Complex Lesion Diagnosis Based on Deep Fusion Networks

Multimodal medical imaging, which combines spatial and functional information, plays an important role in improving the accuracy of complex disease diagnosis. This study aims to address the diagnostic challenges of complex lesions by designing a deep fusion network that integrates channel attention and multi-scale feature extraction. An end-to-end model was built and tested on two public multimodal datasets: glioma and lung tumors. The experimental results show that, compared with existing multimodal fusion methods, the proposed approach achieves better performance in classification accuracy, area under the receiver operating characteristic curve (ROC-AUC), and Dice coefficient for image segmentation. This method provides a new solution for clinical decision support based on multi-source imaging.