Water correction algorithm to improve the classification accuracy: A near-infrared spectroscopy study

It is known that approximately 80% of blood is composed of water (H2O); we developed an algorithm which compensates the light absorbed by the water to improve the signal strength of oxy- and deoxy-hemoglobin (HbO and HbR) during mental activity. To practically implement the proposed algorithm, a new tri-wavelength LED based NIRS systems is used which is capable of detecting not only the concentration changes of HbO and HbR but also that of H2O. The proposed algorithm is tested using an arithmetic task experiment, performed with five healthy male subjects. By using the linear discriminant analysis (LDA) as classifier we were able to distinguish between the activity and rest periods. The results obtained in this study show that the classification accuracy to differentiate between task and rest period increases after we implied the water correction algorithm.