VertiGo - Pilot project for nystagmus detection via webcam

Background Dizziness is one of the most common symptoms in medicine. Detecting the correct diagnosis is complex. Especially in rural areas extensive diagnosis procedures are not always available. The aim of this study is to detect horizontal nystagmus utilizing a commercially available webcam.

Methods In the feasibility study, 30 healthy volunteers participated in a caloric vestibular examination with 44 °C warm water in both ears and an examination by videonystagmography. Afterwards, a further caloric testing and video recording of nystagmus was performed using a FullHD webcam (n=57). The recorded data was analysed with a developed software which uses computer vision techniques to detect faces, eyes and pupils. An algorithm was designed which detected nystagmus from a sequence of horizontal pupil positions. For each dataset, the algorithm differentiated between the presence or absence of nystagmus using a threshold of at least 3 nystagmus in the same direction. To enable an evaluation of the software, the videos were analysed by experienced ENT specialists without Frenzel glasses.

Results In more than 70 % of cases nystagmus were detected by the software. The classification of the datasets using the suggested threshold achieved an accuracy of 59.64 %. Further, 36.36 % of the data which the software evaluates to contain nystagmus are correctly classified (precision).

Summary In the present study first findings show that nystagmus detection with a commercially available webcam is possible using artificial intelligence. In future, further improvements and tests of the software are necessary to increase its accuracy.

Poster-PDF A-1700.PDF

Publication History

Article published online:
10 June 2020