Detecting positional vertigo using an ensemble of 2D convolutional neural networks

The aim of the work presented here was to develop a system that can automatically identify attacks dizziness occurring in patients suffering from positional vertigo, which occurs when sufferers move their head into certain positions. We used our novel medical device, CAVA, record eye- and head-movement data continually for up 30 days diagnosed with disorder called Benign Paroxysmal Positional Vertigo. Building upon previous work, we describe ensemble five 2D Convolutional Neural Networks, using composite recognition features, including eye-movement three-channel accelerometer data. achieve an F1 score 0.63 across 11-fold cross-fold validation experiment, demonstrating detect few seconds motion provoked within over 100 h normal show outperforms 1D Network approach, classifier is superior each individual networks it contains. also demonstrate features provide improved performance results obtained sources independently.