Figure 2.

Subgrouping difference between human and Artificial Intelligence (AI). The reason why clinical trials have not been successful regardless of a deeper understanding of the pathology underlying sepsis is the diverse pathophysiology of sepsis in humans compared to animal sepsis models which are developed using relatively unified methods. Physicians usually classify septic patients by based on the types of primary disease, organ damage, and severity of systematic conditions that are consistent with clinical impressions. However, it is extremely difficult for even a skilled clinician to classify sepsis based on the limited information typically available at the beginning of treatment: medical history, vital signs, and few blood examinations. Thus, clinical trials could not have targeted specific patients. On the other hands, AI may achieve new classification, which is not transparent at first glance and is difficult for humans to understand, by machine learning. Only patients with immunoparalysis detected by AI should be treated with immune checkpoints inhibitors for successful clinical trials, for example.