Target Detection Using Ternary Classification During a Rapid Serial Visual Presentation Task Using Magnetoencephalography Data

Background: The RSVP paradigm is a high-speed paradigm of BCI applications to identify target stimulus. The target stimuli evoke ERP activity of odd-ball effect, and the ERP can be used to identify the onsets of targets among others. However, the ERPs in single sample are various in latency and length, which makes it difficult to accurately discriminate the targets against their neighbors, the near-nontargets. Although the machine learning methods have been used in ERP detection, the problem has not been well solved. Methods: This study performed RSVP experiment, the materials were natural scene pictures with or without pedestrians, the pictures with pedestrians were used as target pictures. MEG data of 10 subjects were acquired during presentation. To overcome the difficulty of ERP detection against their neighbors, we proposed a simple but novel ternary-classification method to train the classifiers. The new method not only distinguished the target against all other samples, but also further separated the target, near-nontarget and other samples. The SVM and CNN in EEGNet architecture classifiers were used to detect the target pictures. Results: We obtained the fairly high target detection scores using SVM and EEGNet classifiers based on MEG data. The proposed ternary-classification method showed the near-nontarget samples can be discriminated from other samples and the separation significantly increased the ERP detection scores in EEGNet classifier. Moreover, the visualization of the new method also suggested the different underling of SVM and EEGNet classifiers in ERP detection of RSVP experiment. Conclusion: In RSVP experiment, the near-nontarget samples contain separatable ERP activity. By separating the nontarget into near- and far-set, we can increase the ERP detection scores using classifiers of CNN model.