Deep Learning for Accurate Diagnosis of Benign Paroxysmal Positional Vertigo

Abstract

Benign paroxysmal positional vertigo (BPPV) is characterized by paroxysms of vertigo and nystagmus triggered by head position changes. The diagnosis of BPPV can be objectively determined through the objective analysis of nystagmus, making it a promising approach towards artificial intelligence (AI) -assisted diagnosis. The diagnostic criteria for BPPV have been clearly defined, and standardized protocols for data collection have been established. Video-oculography utilizing infrared cameras has been employed for the quantification of nystagmus. These objective data can be used to train AI algorithms. Utilizing deep learning models allows for accurate tracking of pupil movement trajectories, facilitating the identification of nystagmus types, and making automated diagnosis of BPPV possible. This chapter summarizes the recent advances in AI-assisted diagnosis of BPPV and discusses the limitations and challenges in clinical practice.