Multi-centre comparison of five eye movement detection algorithms

Although various investigators have suggested algorithms for the automatic detection of eye movements during sleep, objective comparisons of the proposed methods have previously been difficult due to different recording arrangements of different investigators. In this study the results of five eye movement detection algorithms applied to the same data were compared to visually scored data. The percentages of true and false detections are given for various threshold levels in rapid and slow eye movement detections. The methods gave best results when they were used with the same electrode montage they were designed for but the performance decreased when other montages were used. Subtracting the cross-talk of EEG delta activity improved the correctness of eye movement detections.