Exploring Post-Concussion Vestibular Disorders A Retrospective Analysis Using Machine Learning Approaches

Main Article Content

Ignacio Novoa Cornejo https://orcid.org/0000-0002-5071-1332
Victor Mercado Martinez https://orcid.org/0000-0001-8026-8370

Keywords

aprendizaje automático, conmoción encefálica, síntomas vestibulares

Resumen

Introduction: Post-concussion vestibular disorders significantly impact patients' quality of life, but their complex nature challenges traditional clinical assessments.


Aim: To employ machine learning techniques to analyze vestibular disorders in post-concussion patients and describe patient behavior in the otoneurological field.


Material and Methods: This retrospective study examined 75 post-concussion patients in Chile. Random Forest, XGBoost, and Support Vector Regression (SVR) models explored relationships between clinical characteristics and symptom duration. Data included demographic information, concussion details, symptom characteristics, and otoneurological examination results.


Results: SVR demonstrated superior performance (RMSE 151.24), followed by XGBoost (RMSE 224.06) and Random Forest (RMSE 407.99). Key predictors included general health status, sex, and specific vestibular conditions. Vestibulovisual symptoms and specific types of benign paroxysmal positional vertigo (BPPV) emerged as significant factors. Bilateral vestibular hypofunction (BVH) alone did not significantly affect symptom duration. The analysis revealed complex interactions between clinical features and recovery time.


Conclusion: These findings provide insights into the multifaceted nature of post-concussion vestibular disorders and highlight the potential of machine learning in enhancing our understanding of patient trajectories. Results suggest the need for comprehensive evaluation and individualized treatment approaches, potentially leading to improved risk stratification and more targeted interventions for patients with post-concussion vestibular disorders.

Abstract 60 | PDF Downloads 27