Dynamic Spectrum and BP Neural Network for Non-invasive Hemoglobin Measurement

To minimize and hopefully to eliminate the discrepancies among the individuals and the complicated conditions during non-invasive hemoglobin measuring by near-infrared spectroscopy, the Dynamic Spectrum (DS) method was applied. DS is more accurate than the traditional method in hemoglobin non-invasively measurement, which is proved by the theoretical derivation. In vivo measurements were carried out in 60 healthy volunteers, and Back Propagation Neural Network (BP-NN) was used to establish the calibration model of hemoglobin concentration against DS data, which were preprocessed by some special algorithms. The correlation coefficient of the predicted values and the true values was 0.907, which showed that DS method can be applied as a new approach to non-invasive hemoglobin analysis by near-infrared spectroscopy.