Speed of Rapid Serial Visual Presentation of Pictures, Numbers and Words Affects Event-Related Potential-Based Detection Accuracy

• Published in: IEEE transactions on neural systems and rehabilitation engineering Vol. 28; no. 1; pp. 113 - 122
• Main Authors: Lees, Stephanie, McCullagh, Paul, Payne, Phillippa, Maguire, Liam, Lotte, Fabien, Coyle, Damien
• Format: Journal Article
• Language: English
• Published: United States IEEE 01.01.2020; The Institute of Electrical and Electronics Engineers, Inc. (IEEE); IEEE Institute of Electrical and Electronics Engineers
• Subjects: Accuracy; Artificial Intelligence; BCI; Biotechnology; brain-computer interface; Cognitive science; Computer Science; EEG; Electrodes; Electroencephalography; Emerging Technologies; Event-related potentials; Feature extraction; Human-Computer Interaction; Human-computer interface; Image classification; Image detection; Interfaces; Machine Learning; Operators (mathematics); Pictures; Rapid serial visual presentation; Target detection; Target recognition; Testing; Training; Variance analysis; Visualization; BCI; Electroencephalography; Event Related Potentials; Brain- Computer Interface; Rapid Serial Visual Presentation; EEG
• ISSN: 1534-4320; 1558-0210; 1558-0210
• DOI: 10.1109/TNSRE.2019.2953975

Rapid serial visual presentation (RSVP) based brain-computer interfaces (BCIs) can detect target images among a continuous stream of rapidly presented images, by classifying a viewer's event related potentials (ERPs) associated with the target and non-targets images. Whilst the majority of RSVP-BCI studies to date have concentrated on the identification of a single type of image, namely pictures, here we study the capability of RSVP-BCI to detect three different target image types: pictures, numbers and words. The impact of presentation duration (speed) i.e., 100-200ms (5-10Hz), 200-300ms (3.3-5Hz) or 300-400ms (2.5-3.3Hz), is also investigated. 2-way repeated measure ANOVA on accuracies of detecting targets from non-target stimuli (ratio 1:9) measured via area under the receiver operator characteristics curve (AUC) for {N}={15} subjects revealed a significant effect of factor Stimulus-Type (pictures, numbers, words) (F (2,28) = 7.243, {p} = {0.003} ) and for Stimulus-Duration (F (2,28) = 5.591, p = 0.011). Furthermore, there is an interaction between stimulus type and duration: F (4,56) = 4.419, {p} = {0.004} ). The results indicate that when designing RSVP-BCI paradigms, the content of the images and the rate at which images are presented impact on the accuracy of detection and hence these parameters are key experimental variables in protocol design and applications, which apply RSVP for multimodal image datasets.