Using a generalized neural network to identify airway obstructions in anesthetized patients post-operatively based on photoplethysmography

Photoplethysmography has been recently studied as a non-invasive indicator of circulatory and respiratory function. In this study, photoplethysmographic (PPG) data were recorded from patients under the influence of anesthesia, but not intubated. Both time and frequency domain features were extracted from the PPG and used as inputs to a neural network classifier. This classifier considers inter-subject variability so that it generalizes well to a large population. This classifier provided 86.1% accuracy to classify segments as being times of ‘obstructed’ vs. ‘normal’ airways status.