A machine learning algorithm was developed from high fidelity arterial pressure waveforms. These waveforms contain immense information in the morphology of various phases of the cardiac cycle. Features such as signal features, Flotrac and COtrek features, complexity, variability, spectral features, delta change and baroreceptor features are all incorporated and the resulting machine learning algorithm could predict hypotension 15 mins earlier than clinically apparent in the surgical setting.
The future may see non-invasive pulse oximeter based analysis without the Flotrac proprietary technology used here. At any rate hypotension with a MAP < 65 mm Hg is now recognized as a potentially very detrimental outcome predictor which emphasizes the importance of this study.