Addressing Structural Bias in Interpretation of Pulse Oximetry