A predictive model for early neurological deterioration in medullary infarction based on explainable machine...

Area of Science:

• Neurology
• Artificial Intelligence
• Medical Informatics

Background:

• Medullary infarction is a severe ischemic stroke subtype.
• Early neurological deterioration (END) significantly impacts patient outcomes.

Purpose of the Study:

• Develop a machine learning model to predict END risk in medullary infarction patients.
• Provide a clinical decision support tool for timely intervention.

Main Methods:

• Multicenter retrospective study of 352 medullary infarction patients.
• Collected multi-dimensional clinical data.
• Developed and evaluated five machine learning models, including a single-layer neural network and Naïve Bayes classifier, using AUC, precision-recall, and calibration curves.

Main Results:

• The single-layer neural network achieved an AUC of 0.873 for END prediction.
• The model demonstrated favorable calibration and overall performance.
• The Naïve Bayes model showed excellent classification consistency.

Conclusions:

• A highly accurate single-layer neural network and a robust Naïve Bayes classifier were developed.
• These models provide an interpretable tool for assessing END risk in stroke patients.
• The findings support clinical application across diverse settings.