Related Experiment Video
Updated: Jan 9, 2026
Autologous Blood Injection to Model Spontaneous Intracerebral Hemorrhage in Mice
Published on: August 24, 2011
Machine learning-based prediction of acute kidney injury after intracerebral hemorrhage: Comparison of multiple model
Junzhang Huang1, Ningkun Xiao2,3
1Department of General Surgery, Lianjiang Traditional Chinese Medicine Hospital, Lianjiang City, Zhanjiang, Guangdong Province, China.
To identify key risk factors for acute kidney injury (AKI) in patients with intracerebral hemorrhage (ICH) using bibliometric analysis and machine learning, and to explore the mediating role of hemoglobin (Hb) in the association between hypertension and AKI. A bibliometric analysis of English-language publications from 2014 to 2024 was conducted to evaluate research trends and thematic clusters related to ICH and AKI. Clinical data from 944 ICH patients in the Medical Information Mart for Intensive Care-IV (MIMIC-IV) database were analyzed, including 46 patients who developed AKI. A total of 120 machine learning model combinations were developed and compared using area under the receiver operating characteristic curve (AUC). SHapley Additive exPlanations (SHAP) and clustering heatmaps were employed to interpret feature importance and patient-level heterogeneity. Mediation analysis was used to assess the indirect effect of Hb on the relationship between hypertension and AKI. A total of 122 relevant publications were identified, revealing a growing academic focus on integrated neurocritical care and the emergence of stable interdisciplinary research clusters. Among the machine learning models, the Random Forest (RF) + NaiveBayes + Tree ensemble achieved the highest predictive performance (AUC = 0.764). SHAP analysis identified platelet count (PLT) as the most influential predictor of AKI risk. Clustering analysis revealed substantial heterogeneity in feature contributions across patient subgroups. Mediation analysis confirmed that Hb significantly and negatively mediated the effect of hypertension on AKI occurrence (P < .05). The RF + NaiveBayes + Tree ensemble demonstrated superior predictive power for early AKI risk identification. SHAP analysis highlighted PLT as a key predictor, with notable inter-individual variability in risk profiles. The observed mediating effect of Hb between hypertension and AKI offers exploratory mechanistic insights that warrant further validation. While these findings contribute to the understanding of ICH-associated AKI, the current model remains exploratory and is not yet ready for direct clinical implementation. Future studies with larger, multicenter cohorts are needed to refine the model and assess its clinical utility.
To identify key risk factors for acute kidney injury (AKI) in patients with intracerebral hemorrhage (ICH) using bibliometric analysis and machine learning, and to explore the mediating role of hemoglobin (Hb) in the association between hypertension and AKI. A bibliometric analysis of English-language publications from 2014 to 2024 was conducted to evaluate research trends and thematic clusters related to ICH and AKI. Clinical data from 944 ICH patients in the Medical Information Mart for Intensive Care-IV (MIMIC-IV) database were analyzed, including 46 patients who developed AKI. A total of 120 machine learning model combinations were developed and compared using area under the receiver operating characteristic curve (AUC). SHapley Additive exPlanations (SHAP) and clustering heatmaps were employed to interpret feature importance and patient-level heterogeneity. Mediation analysis was used to assess the indirect effect of Hb on the relationship between hypertension and AKI. A total of 122 relevant publications were identified, revealing a growing academic focus on integrated neurocritical care and the emergence of stable interdisciplinary research clusters. Among the machine learning models, the Random Forest (RF) + NaiveBayes + Tree ensemble achieved the highest predictive performance (AUC = 0.764). SHAP analysis identified platelet count (PLT) as the most influential predictor of AKI risk. Clustering analysis revealed substantial heterogeneity in feature contributions across patient subgroups. Mediation analysis confirmed that Hb significantly and negatively mediated the effect of hypertension on AKI occurrence (P < .05). The RF + NaiveBayes + Tree ensemble demonstrated superior predictive power for early AKI risk identification. SHAP analysis highlighted PLT as a key predictor, with notable inter-individual variability in risk profiles. The observed mediating effect of Hb between hypertension and AKI offers exploratory mechanistic insights that warrant further validation. While these findings contribute to the understanding of ICH-associated AKI, the current model remains exploratory and is not yet ready for direct clinical implementation. Future studies with larger, multicenter cohorts are needed to refine the model and assess its clinical utility.
