Clinical Decision Support for Vestibular Diagnosis: Large-Scale Machine Learning with Lived Experience Coaching

Pastor et al. (2025) developed a machine learning system (MLS) to help make a vestibular diagnosis based on patient symptoms. They utilized diagnostic data from 3,349 patients who were initially evaluated through the Seoul National University Hospital to create their machine learning program. The researchers sought to create a model that aids in the efficacy and accuracy of a vestibular diagnosis.

Patients using this system must answer 145 history questions. They limited their MLS to differentiating between vestibular migraine (VM), persistent postural perceptual dizziness (PPPD), benign paroxysmal positional vertigo (BPPV), Meniere’s disease (MD), and hemodynamic orthostatic dizziness (HOD). The accuracy of this model was greater than 85 percent in diagnosing a vestibular disorder, and it did a particularly good job (greater than 90 percent) in identifying MD, PPPD, and HOD.

Reference

Callejas Pastor, C.A., Ryu, H.T., Joo, J.S. et al. (2025). Clinical decision support for vestibular diagnosis: large-scale machine learning with lived experience coaching. NPJ Digital Medicine, 8, 487.