JI-ADF: A Trimodal Deep Learning Framework for Enhanced Skin Lesion Classification
JI-ADF is a trimodal deep learning framework for skin lesion classification that overcomes the underutilization of multimodal evidence in existing systems. It integrates dermoscopic images, clinical photographs, and patient metadata using joint representation learning, modality-specific supervision, and an adaptive decision fusion mechanism. The framework also incorporates a multimodal fusion attention module to improve cross-modal reasoning and has shown robust performance on the MILK10k benchmark, enhancing sensitivity and Dice scores while maintaining high specificity and calibration.