Correlation of Cognitive Scores and the Onset of Hypoxia.

The negative effects of hypoxia on human cognitive function have been well documented. In this study we assess the correlation of performance in the SynWin cognitive Multi-Task Battery (MTB) and the onset of hypoxia and describe the use of cognitive assessment scores for real-time hypoxia detection. We performed a correlation analysis between MTB scores (Arithmetic, Memory, Audio Monitoring, Video Monitoring tasks) and blood oxygen saturation levels to discover if the scores are good candidates to detect hypoxia. Since this analysis showed positive correlation, we proceeded to develop a parallel decision fusion system that uses these cognitive scores for real-time hypoxia detection using the Neyman-Pearson criterion. We demonstrate that MTB scores have considerable hypoxia detection potential and can be used (if measurable passively) in a real-time detection framework. Analysis of receiver operating characteristic (ROC) curves established a hierarchy of importance of the various MTB modules. The Arithmetic task module had the most significant contribution toward correct hypoxia detection (improvement of ∼13.5% and ∼13.9% in detection accuracy under global false alarms of 0.1 and 0.05, respectively), followed by the Memory and Audio Monitoring modules. Fusion of multiple cognitive assessment scores resulted in significantly higher detection accuracy (86%) than using any one of the scores by itself. When available, cognitive assessment scores can be a useful tool for real-time hypoxia detection. Since these MTB tests also assess neuropsychological functioning, study of distributed detection systems based on MTB scores could help in designing tests that are more useful for detecting hypoxic symptoms.