Common Spatio-Temporal Patterns in Single-Trial Detection of Event-Related Potential for Rapid Image Triage

Searching for targets in large volume imagery is a challenging problem. A potentially promising solution, rapid image triage (RAIT), is essentially a cortically-coupled computer vision technique based on single-trial detection of event-related potentials (ERP). This paper presents a novel feature selection method, termed common spatio-temporal patterns (CSTP), which is critical for robust single-trial detection of ERP. Unlike past methods such as conventional common spatial pattern (CSP) method whereby only spatial patterns of ERP are considered, the proposed method exploits both spatial and temporal patterns of ERP, providing complementary spatial and temporal features for high accurate single-trial ERP detection. Numerical study using data collected from 20 subjects demonstrates that the proposed method offers significant performance improvement over conventional CSP method (p-value < 0.01) for rapid image triage.