Using multiple machine learning algorithms to predict spinal cord injury in patients with cervical spondylosis: a...

Area of Science:

• Neurology
• Medical Informatics
• Orthopedic Surgery

Background:

• Degenerative cervical spondylosis is a progressive condition with significant global health implications.
• Spinal cord injury is a severe complication of cervical spondylosis, necessitating early detection and intervention.
• Machine learning (ML) offers powerful tools for analyzing complex medical data and predicting disease trajectories.

Purpose of the Study:

• To develop and validate a machine learning model for predicting spinal cord injury in patients with degenerative cervical spondylosis.
• To identify key predictive factors contributing to spinal cord injury development in this patient population.
• To support clinical decision-making and personalized treatment strategies for degenerative cervical spondylosis.

Main Methods:

• Retrospective analysis of clinical data from 737 patients across three hospitals.
• Development and comparison of 10 machine learning models, including random forest.
• Identification of 11 core predictive factors using univariate analysis and LASSO regression.
• Model validation using training, testing, and external validation sets.

Main Results:

• The random forest model demonstrated superior predictive performance with high AUC values and accuracy.
• Eleven core predictive factors for spinal cord injury were identified.
• The study observed a high incidence of cervical spondylosis, with a concerning trend in younger individuals.

Conclusions:

• Machine learning, particularly the random forest model, effectively predicts spinal cord injury risk in degenerative cervical spondylosis.
• The developed model can assist surgeons in creating precise, individualized treatment plans.
• Early prediction and tailored interventions can enhance therapeutic effectiveness and reduce unnecessary procedures.