Predicting Early Treatment Effectiveness in Bell's Palsy Using Machine Learning: A Focus on Corticosteroids and
Jheng-Ting Luo1, Yung-Chun Hung1,2, Gina Jinna Chen3
1In-Service Master Program in Artificial Intelligence in Medicine, College of Medicine, Taipei Medical University, Taipei, Taiwan.
International Journal of General Medicine
|November 14, 2024
Summary
Machine learning models can predict Bell's palsy recovery. Early prednisolone treatment and patient age are key factors for better outcomes in facial nerve paralysis.
Area of Science:
- Neurology
- Medical Informatics
Background:
- Bell's palsy, a form of facial nerve paralysis, often results in unilateral facial weakness.
- While many patients recover spontaneously within months, a significant portion experiences incomplete recovery.
Purpose of the Study:
- To investigate the effectiveness of early treatment for Bell's palsy using machine learning.
- To compare the predictive performance of six machine learning models for recovery outcomes.
Main Methods:
- Data from 493 patients across 17 hospitals were analyzed.
- Patients received Prednisolone, Acyclovir, both, or placebo, with outcomes assessed via the House-Brackmann scale.
- Six machine learning models were employed to predict recovery, with performance evaluated using AUC, precision, recall, and F1-score.
Main Results:
- Overall recovery rates were 72.6% at 3 months and 89.5% at 9 months.
- Logistic regression achieved the highest predictive performance (AUC = 0.751 at 3 months, AUC = 0.720 at 9 months).
- Patient age and Prednisolone administration were identified as the most significant predictors of recovery.
Conclusions:
- Machine learning models show promise in predicting Bell's palsy treatment effectiveness.
- Logistic regression demonstrated superior predictive capabilities for recovery.
- The study highlights the novel significance of patient age in prognosis and suggests developing age-specific models for tailored treatment strategies.
