Construction of an Interpretable Model of the Risk of Post-Traumatic Brain Infarction Based on Machine Learning...

Area of Science:

• Neuroscience
• Medical Informatics
• Clinical Prediction Modeling

Background:

• Post-traumatic cerebral infarction (PTCI) is a severe complication of traumatic brain injury (TBI).
• PTCI can lead to permanent neurological deficits or mortality.
• Identifying PTCI predictors and developing predictive models are critical for clinical management.

Purpose of the Study:

• To investigate factors associated with PTCI in TBI patients.
• To develop and compare machine learning (ML) models for PTCI risk prediction.
• To utilize SHAP values for interpreting ML model predictions.

Main Methods:

• Retrospective analysis of clinical data from 1484 TBI patients.
• Identification of predictive factors using LASSO and multivariable logistic regression.
• Development and comparison of ML classification models, with SHAP value interpretation.

Main Results:

• Key predictors for PTCI included age, bilateral brain contusions, platelet count, uric acid, glucose, traumatic subarachnoid hemorrhage, and surgical treatment.
• The logistic regression (LR) model demonstrated superior performance with an AUC of 0.821 and accuracy of 0.845.
• The LR model exhibited stable performance during ten-fold cross-validation.

Conclusions:

• ML algorithms effectively predict PTCI risk by integrating demographic and clinical factors.
• SHAP value interpretation provides insights for personalized patient treatment strategies.
• These models bridge the gap between complex clinical data and actionable clinical insights.