Can a smartphone app with AI replace the VNG in caloric testing?

Introduction Dizziness is one of the most common symptoms in medicine, but the diagnosis is very complex and depends on the examiner"s expertise. In addition, this experience and expensive, equipment-based, error-prone diagnostics are not available across the country.

Methods An Android smartphone app was developed that can be used by oneself or third parties for nystagmography. VNG was carried out on 19 healthy volunteers at rest and after caloric testing of each ear in a conventional manner and using the smartphone app. In addition, a KIT was performed with the VNG. The app was used and evaluated directly on the smartphone without any additional hardware. The purpose of the self-developed app is to check whether the VNG-App on the smartphone using AI (eye tracking) can replace the VNG gold standard with infrared technology.

Results Taking into account a limit of 6 nystagmus in the same direction, nystagmus could be detected in 72% of the recordings. A sensitivity of 81%, PPV of 67% and specificity of 42% were achieved. An average SPV of 14°/s (±0.2 SEM) vs. 12.4°/s (±0.8) and frequency of 16 (±0.7) in 20 s vs. 36.9 (±3.2) in 30 s was achieved in the VNG app group and gold standard VNG group. When comparing both techniques with regard to the individual SPV per side, a small bias of 0.9 was observed. A total of 6 subjects were measured with vestibular hypofunction according to the SPV in the gold standard, which was not confirmed by the mobile VNG app or the KIT.

Discussion Our findings show that a mobile, cost-effective and modified VNG using AI is feasible and that it can approximate the common VNG in the event of failure or absence. Limitations of the VNG app are the lack of fixation suppression as well as light dimming and the frame rate of 30 vs. 120 Hz.

Funding informations BMBF

Publication History

Article published online:
19 April 2024