Table 1.
Key features and application scenarios of the datasets.
Fig 1.
Architecture of the MCoder-T model for multimodal medical coding and intelligent classification.
Fig 2.
Architecture of the causal masked attention mechanism.
Fig 3.
Architecture of the fine-grained cross-modal fusion module using PCMA.
Fig 4.
Architecture of the multi-task learning and LoRA integration module.
Table 2.
Performance comparison under standard and robustness settings on COVID-19 and MedMNIST datasets.
Fig 5.
Performance comparison of MCoder-T and baseline models under standard settings.
Fig 6.
Comparison of experimental results for MCoder-T and baseline models.
Table 3.
Ablation study results of MCoder-T model by removing individual modules.
Fig 7.
Effect of MCoder-T model after module ablation.
Table 4.
Ablation study results of MCoder-T model by removing multiple modules.
Fig 8.
Ablation study results of multiple modules in MCoder-T model for investigating the interaction of module combinations.