Sepsis is a leading cause of acute kidney injury (AKI), with nearly 50% of sepsis patients progressing to acute kidney disease (AKD). In the long term, many cases deteriorate into chronic kidney disease (CKD), potentially requiring dialysis treatment. Current clinical diagnostics rely primarily on serum creatinine levels, estimated glomerular filtration rate (eGFR), and urine output, yet these biomarkers often lack sensitivity and specificity, making early risk assessment challenging.
To address this clinical gap, a research team led by Professor Mei-Yi Wu at the Ministry of Health and Welfare’s Shuang Ho Hospital has developed an AI-driven immune biomarker prediction system. By integrating immune cell phenotyping with machine learning algorithms, this innovative tool enables early identification of AKI risk in sepsis patients, allowing physicians to initiate timely interventions and potentially improving kidney function recovery and dialysis-free survival rates.
Professor Wu’s team discovered that immune inflammation plays a pivotal role in AKI progression, suggesting that tracking immune cell dynamics could enhance early detection. Using flow cytometry, the researchers analyzed peripheral blood mononuclear cells (PBMCs) from sepsis patients, employing advanced high-dimensional cytometry data analysis.
Through unsupervised clustering (FlowSOM) and semi-supervised machine learning methods (ACDC, LDA), they identified key immune cell subsets—such as regulatory T cells (Tregs) and NK CD56d cells—strongly associated with AKI development. Leveraging decision tree algorithms, the team integrated immune cell phenotypic data with biochemical markers (e.g., serum creatinine and blood urea nitrogen) to construct a real-time risk assessment model.
Clinical validation demonstrated that this AI-based system outperforms traditional diagnostic methods, enabling earlier detection of disease progression and prompt clinical intervention, ultimately improving patient outcomes. Moving forward, the team plans to incorporate in vitro diagnostic (IVD) technologies, further enhancing the system’s clinical accessibility and usability.
According to Professor Wu, this predictive model not only reduces sepsis-associated AKI incidence and mortality but also holds promise for broader applications in high-risk kidney disease populations, reinforcing the development of precision medicine. Future efforts will focus on scaling up clinical validation, integrating the system into standard sepsis care protocols, and collaborating with biotech companies to develop a fully automated diagnostic platform.
By bridging AI, immunology, and nephrology, this innovation is set to enhance medical efficiency, alleviate healthcare burdens, and ultimately improve patient quality of life.
Resource: 敗血症患者要謹「腎」 AI分析免疫細胞預測惡化風險
