VERTIGO IDENTIFICATION AND PREDICTION

Abstract

Background: Vertigo is a common and often debilitating symptom reflecting disorders of the vestibular system. While clinical identification of vertigo causes (such as peripheral vs. central vestibular dysfunction) is well-developed, there remains a gap in predicting episodes of vertigo, particularly in recurrent conditions like Benign Paroxysmal Positional Vertigo (BPPV) or Menière’s disease. Advances in sensor technologies and machine-learning offer new opportunities for forecasting vertigo risk. Objective: This study aims to (1) develop a systematic clinical workflow for the identification of vertigo etiology, and (2) build and validate a predictive model that forecasts short-term risk of vertigo episodes using a combination of clinical, sensor-derived and environmental data.then applies bio-inspired algorithms to identify the most significant features influencing heart disease risk. Selected features are subsequently fed into classification models, including Support Vector Machines (SVM), Decision Trees, and Neural Networks, to predict the likelihood of heart disease. Experimental results demonstrate that bio-inspired algorithms improve feature selection efficiency, reduce computational complexity, and enhance the overall accuracy, sensitivity, and specificity of heart disease prediction models. This approach provides a robust decision-support tool for medical professionals, enabling timely diagnosis and improved patient care