Prediction of respiratory measurements based on cross embedding techniques

Measurements of multiple physiological signals are required for diagnostic procedures such as for sleep disordered breathing. Accuracy of automated diagnostic procedures and home based screening methods can be affected when phisiological measurements contains artifacts or signal losses. We investigate on predicting one physiological signal measurement from others, using dependencies exists in physiological signals, in order to obtain a measure of reliability to the measurements. Modeling such relationships are done with the use of artificial neural networks. We conclude that via such cross prediction tasks, it is possible to identify and correct both artifacts and signal losses in these measurements.