Abstract
Purpose :
This study investigates the efficacy of an advanced artificial intelligence (AI)-powered clinical decision support system specifically designed for the detection of Nystagmus. The system, leveraging real-time clinical data, aims to revolutionize the diagnosis of Nystagmus and explore its potential for integration into telemedicine platforms
Methods :
We developed a cloud-based deep learning (DL) application capable of real-time tracking of eye movements and identifying 468 facial landmarks to diagnose Nystagmus in vertigo patients. Patients submitted self-recorded videos of their eye movements, which were transformed into data points and analyzed by the DL software. The software computed the slow-phase velocity (SPV) values from eye movement data, represented graphically. These SPV values were cross-verified against standard video-nystagmography (VNG) readings and clinician assessments. This procedure was repeated threefold for each of the ten participants. The DL software's outcomes were statistically compared to the VNG results, with a significance threshold set at p < 0.05.
Results :
The analysis of the data revealed significant outcomes: A) The results showed statistical significance with a value of p < 0.05; B) The mean square error (MSE) was calculated to be 0.00459; C) The mean deviation between the SPV values from the DL software and those obtained from VNG readings was ±4.8%, illustrating the high accuracy and reliability of the DL model in detecting Nystagmus.
Conclusions :
This innovative, clinician-guided DL software shows promising potential as a feasible alternative to traditional in-office consultations for diagnosing Vertigo and associated Nystagmus. It is particularly applicable for conditions like Benign Paroxysmal Positional Vertigo (BPPV), commonly treated with canalith repositioning procedures. The software’s ability to monitor changes in Nystagmus before and after treatment highlights its clinical value, underscoring the need for further research and exploration in this field. The integration of such AI-based systems into telemedicine could significantly enhance patient care, offering a more flexible, cost-effective, and accessible approach to managing vertigo and related conditions.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.