High-resolution LA-ICP-MS imaging combined with attention-enhanced residual network reveal cadmium-induced metal distribution heterogeneity and toxicity in Mytilus galloprovincialis.

Marine bivalves often face heavy metal stress, yet we still lack of high-resolution way to observe the metal pollution status in soft tissue. This study provides the first high-resolution spatial mapping of cadmium (Cd) and associated heavy metals in marine environmental monitor Mytilus galloprovincialis, revealing distinct organ- and cell-specific accumulation, with the digestive gland, especially digestive cells, being the primary site of Cd uptake. Cd co-localizes with several metals, including calcium, chromium, copper, and zinc, suggesting shared uptake pathways and potential synergistic toxicity. Competitive and cooperative interactions between Cd and other metals highlight the complexity of metal homeostasis under Cd exposure. To analyze these datasets, an attention-enhanced Residual Network (attention-ResNet) was implemented, incorporating residual blocks, spatial attention modules, and Gaussian noise augmentation to expand limited data (∼100 images per isotope), enabling accurate classification of Cd exposure vs. non- Cd exposure images and detection of subtle inter-elemental interactions. This integrative approach advances ecotoxicological insights and advocates for interdisciplinary methods combining imaging, machine learning, and modeling to elucidate pollutant impacts on marine organisms, providing a foundation for targeted strategies to mitigate heavy metal risks in marine ecosystems.