Machine Learning Predicts Neurosurgical Outcomes Using Liver Function Indicators

Overview

A retrospective study developed a CatBoost machine learning model to predict unfavorable outcomes in neurosurgical patients, integrating neurological and liver function markers. Key predictors included Glasgow Coma Scale score, age, and liver enzymes such as AST, ALKP, and albumin levels, with the model achieving an AUC of 0.932.

Background

Neurosurgical diseases like traumatic brain injury and intracerebral hemorrhage have high mortality and disability rates, making early prognostic assessment critical. The liver-brain axis, involving interactions between hepatic dysfunction and neurological outcomes, has emerged as an important research focus. Liver function markers such as AST, ALKP, and albumin have been linked to neurological deterioration but remain underexplored across diverse neurosurgical populations. Machine learning offers advantages over traditional models by capturing complex, non-linear relationships between these markers and patient outcomes.

Data Highlights

Key Findings

• The CatBoost model demonstrated excellent predictive performance for unfavorable neurosurgical outcomes (AUC=0.932).
• Lower Glasgow Coma Scale (GCS) scores at admission and older age were strongly associated with poor prognosis.
• Elevated mean AST and ALKP levels, including initial ALKP, correlated with increased risk of unfavorable outcomes.
• Lower mean and minimum albumin levels were linked to worse functional status at discharge.
• Integrating liver function markers with neurological variables enhanced model accuracy beyond traditional clinical predictors.
• Shapley additive explanations provided interpretability of feature contributions to model predictions.

Clinical Implications

Incorporating liver function tests such as AST, ALKP, and albumin into early prognostic assessments can improve identification of neurosurgical patients at risk for poor outcomes. Machine learning models like CatBoost can effectively integrate these markers with neurological data to guide clinical decision-making. Routine evaluation of liver function may facilitate timely interventions and personalized treatment strategies.

Conclusion

This study highlights the prognostic value of liver function indicators in neurosurgical patients and demonstrates that machine learning models integrating these markers with neurological assessments can accurately predict adverse outcomes. Further multicenter validation is warranted to confirm these findings and explore underlying mechanisms.

References

1. Tianjin Medical University General Hospital Study 2023 -- Predicting Adverse Outcomes in Neurosurgical Patients Using Machine Learning