Sepsis-Induced Acute Respiratory Distress Syndrome (SI-ARDS) presents significant diagnostic and prognostic challenges due to its complex clinical manifestations and high mortality rate. We developed a deep Knowledge-Driven multi-Modal Fusion (KDMF) framework for the accurate diagnosis and prognosis of SI-ARDS. The model leverages multi-modal data, including CT images, CT reports, and laboratory indicators, alongside a disease-specific knowledge graph. KDMF achieves superior performance in predicting SI-ARDS incidence (AUC 0.930) and time to 28-day mortality (AUC 0.843, C-index 0.833). Comprehensive error analysis and ablation studies demonstrate the critical contributions of each data modality and the integrated knowledge graph. The results highlight the potential of KDMF to enhance early intervention and treatment strategies, underscoring the robustness and interpretability of the framework in clinical applications. Sepsis is a life-threatening condition that can lead to a serious lung injury called ARDS, which is difficult to diagnose early and has a high risk of death. This study developed a computer model called KDMF to help medical practitioners identify sepsis patients who are likely to develop ARDS and predict their risk of death within 28 days. The model combines multiple types of patient data—including scans of the lungs, text from radiology reports, and routine lab results—along with medical knowledge built into the system. When tested on real patient data, the model was highly accurate at predicting both the onset of ARDS and patient survival. It also helped explain which factors, such as specific symptoms or lab values, were most important for its predictions. This tool could support doctors in making faster and more informed decisions, potentially improving treatment and outcomes for high-risk patients in the intensive care unit. Chen, Gu, Zhang, Zhang, et al. developed a computer model called KDMF to help doctors identify sepsis patients who are likely to develop ARDS and predict their risk of death within 28 days. The model, combining multiple types of patient data, was highly accurate at predicting both the onset of ARDS and patient survival.
Chen et al. (Thu,) studied this question.