Automated CT-derived visceral-to-subcutaneous fat ratio as a prognostic imaging biomarker for mortality in sepsis.
Authors
Affiliations (7)
Affiliations (7)
- Nurturing Center of Jiangsu Province for State Laboratory of AI Imaging & Interventional Radiology, Department of Radiology, Zhongda Hospital, School of Medicine, Southeast University, Nanjing, China.
- Jiangsu Provincial Key Laboratory of Critical Care Medicine, Department of Critical Care Medicine, Zhongda Hospital, School of Medicine, Southeast University, Nanjing, China.
- The First People's Hospital of Lianyungang, The Affiliated Lianyungang Hospital of Xuzhou Medical University, Lianyungang, China.
- Jiangsu Provincial Joint International Research Laboratory of Medical Information Processing, School of Computer Science and Engineering, Southeast University, Nanjing, China.
- Key Laboratory of New Generation Artificial Intelligence Technology and Its Interdisciplinary Applications (Southeast University), Ministry of Education, Nanjing, China.
- Department of Medical Imaging, Subei People's Hospital, Medical School of Yangzhou University, Yangzhou, China.
- Nurturing Center of Jiangsu Province for State Laboratory of AI Imaging & Interventional Radiology, Department of Radiology, Zhongda Hospital, School of Medicine, Southeast University, Nanjing, China. [email protected].
Abstract
To evaluate the prognostic value of automated CT-derived visceral-to-subcutaneous fat ratio (VSR) and its associations with inflammatory and metabolic biomarkers in sepsis. In this multicenter study, 1716 patients with sepsis were assigned to training (n = 905), internal validation (n = 388), and external validation (n = 423) cohorts. Body composition was segmented using nnU-Net. Five machine learning-based survival models were developed: gradient boosting survival, fast kernel survival support vector machine, extra survival trees (EST), random survival forest, and Cox proportional hazards survival model. Model performance was assessed using the C-index and time-dependent area under the curve (AUC). Cox models were used to identify prognostic factors and evaluate interactions between VSR and biomarkers. The incremental value of VSR was assessed by comparing clinical models with and without VSR. The median age was 69 years (IQR: 58-77), and 68.2% were male. EST was selected based on internal validation performance, with a C-index of 0.845 and a mean time-dependent AUC of 0.905. Multivariable Cox analysis confirmed VSR as an independent prognostic factor (HR = 2.117, 95% CI: 1.946-2.301, p < 0.001). Significant interactions were observed between VSR and lactic acid (p = 0.008) and procalcitonin (p = 0.026). Adding VSR increased the clinical model's C-index by 0.049, 0.080, and 0.020 in the training, internal validation, and external validation cohorts, respectively. Higher VSR levels were associated with increased mortality risk. VSR interacted with procalcitonin and lactic acid and provided incremental value beyond clinical variables. Question Body composition parameters are important in sepsis prognosis, but the relationship of VSR with mortality and inflammatory and metabolic markers remains unclear. Findings CT-derived VSR remained independently associated with mortality after adjustment, added prognostic value beyond clinical variables, and interacted with procalcitonin and lactic acid. Clinical relevance CT-derived VSR offers a simple imaging marker that improves risk assessment beyond clinical variables and facilitates the identification of high-risk patients with sepsis.