VertiGo-App – Smartphone-based video nystagmography using artificial intelligence

Dizziness is one of the most common medical symptoms. The diagnosis is complex, expensive and not always available across the country. In order to improve this, a location and time-independent solution is to be developed. The aim of this study was to perform nystagmus detection with a smartphone and valuate it using artificial intelligence (AI).

In the feasibility study, 7 healthy subjects underwent caloric examinations. The video nystagmography (VNG) was performed using the selfdesigned prototype of the VertiGo app. Reacordings were made with (n = 20) and without a surgical mask (n = 35) independently in selfie mode or by external operation of the rear camera. The videos were analyzed with the selfdesigned software that uses machine learning. An algorithm detected nystagmus in sequences of horizontal pupil positions. A distinction was made between the presence or absence of nystagmus using a threshold of at least two nystagmus in the same direction.

Individual nystagmus were recognized in 85% of the videos. Considering the threshold, a positive predictive value of 94%, a sensitivity of 36% and a specificity of 88% were achieved. Using the selfie or rear camera as well as wearing a surgical mask did not affect the VNG.

In this study, the mobile nystagmus detection was carried out using a smartphone in self-use in selfie mode as well as by another person using the rear camera and was evaluated with a software based on AI. The recording quality was not affected by wearing a surgical mask. The results show improvements in the performance compared to the preliminary studies with a webcam. The current prototype app needs further improvements in order to conduct clinical trials.

gefördert durch das BMBF Mensch-Technik-Interaktion https://www.interaktive-technologien.de/projekte/vertigo

Publication History

Article published online:
24 May 2022