Predicting serious diagnoses in vertigo patients using clinical data and machine learning techniques

Abstract

Vertigo is a condition that results in a sensation of rotating or dizziness, and can arise from one of the following causes – from inflammation of the inner ear to life-threatening conditions like brain stroke, brain tumor, etc. Therefore, Accurate and timely diagnosis are critical to effectively manage and treat patients, preventing potential complications. However, conventional diagnostic procedures depend basically on clinical orientation and may involve numerous tests, which can delay the process of identification of serious conditions. The development of machine learning in recent years provides potential methods for risk assessment in individual vertigo patients during the initial stages. Machine learning models can use extensive data and high-quality algorithms to detect tendencies and factors that can lead to severe diseases by analyzing patients’ symptoms, history, and test results. In their current form, these clinical models offer considerable benefits: they support experienced and skilled clinicians in making better decisions and addressing patients’ needs that require prompt attention. In order to build a robust risk prediction model for this study, we proposed and examined various machine learning models along with a combination of feature selection approaches. When analyzing the results of all considered models, such as boosting and ensemble methods, the best performance was observed by Logistic Regression. Combined with the correlation-based feature selection and splitting ratio of 75:15:10, this model yielded 97% sensitivity with data imbalance being maintained by the SMOTE method. The sensitivity of the model suggested in this study could contribute to upgrade the diagnostic abilities, reducing the load on health resources due to the minimization of ineffective tests, and spotlight the risky cases. This machine learning-driven approach holds the potential to transform the management of vertigo, improving patient outcomes and ensuring timely intervention for those with serious underlying conditions.