Peer Reviewed Journal via three different mandatory reviewing processes, since 2006, and, from September 2020, a fourth mandatory peer-editing has been added.
Searching for points of interest (POI) in large-volume imagery is a challenging problem with few good solutions. In this work, a neural engineering approach called rapid image triage (RIT) which could offer about a ten-fold speed up in POI searching is developed. It is essentially a cortically-coupled computer vision technique, whereby the user is presented bursts of images at a speed of 6–15 images per second and then neural signals called event-related potential (ERP) is used as the ‘cue’ for user seeing images of high relevance likelihood. Compared to past efforts, the implemented system has several unique features: (1) it applies overlapping frames in image chip preparation, to ensure rapid image triage performance; (2) a novel common spatial-temporal pattern (CSTP) algorithm that makes use of both spatial and temporal patterns of ERP topography is proposed for high-accuracy single-trial ERP detection; (3) a weighted version of probabilistic support-vector-machine (SVM) is used to address the inherent unbalanced nature of single-trial ERP detection for RIT. High accuracy, fast learning, and real-time capability of the developed system shown on 20 subjects demonstrate the feasibility of a brainmachine integrated rapid image triage system for fast detection of POI from large-volume imagery.
