Abstract:

OBJECTIVE: To develop and test a method for standardized calibration of pulse oximeters. METHODS: A novel pulse oximeter calibration technique capable of simulating the behavior of real patients is discussed. It is based on an artificial finger with a variable spectral-resolved light attenuator in conjunction with an extensive clinical database of time-resolved optical transmission spectra of patients fingers in the wavelength range 600-1000 nm. The arterial oxygen saturation of the patients at the time of recording was derived by analyzing a corresponding blood sample with a CO-oximeter. These spectra are used to compute the modulation of the light attenuator which is attached to the artificial finger. This calibration method was tested by arbitrarily playing back recorded spectra to pulse oximeters and comparing their display to the value they displayed when the spectra were recorded. RESULTS: We were able to demonstrate that the calibrator could generate physiological signals which are accepted by a pulse oximeter. We also present some experience of playing back recorded patient spectra. The mean difference between the original reading of the pulse oximeters and the display when attached to the calibrator is 1.2 saturation points (displayed oxygen saturation SpO2) with a standard deviation of 1.9 saturation points. CONCLUSIONS: The tests have shown the capabilities of a spectral light modulator for use as a possible calibration standard for pulse oximeters. If some improvements of the current prototype can be achieved we conclude from the experience with the device that this novel concept for the calibration of pulse oximeters is feasible and that it could become an important tool for assessing the performance of pulse oximeters.

Abstract:

Up to now one unsolved challenge in anaesthesia is the assessment of depth of anaesthesia during surgery. No general purpose on-line monitoring system predicting depth or quality of anaesthesia exists. The analysis of spontaneous (EEG) and evoked electrical brain activities (AEP) leads to methods assessing depth of anaesthesia. A monitor concept was developed consisting of the three functional components EEG recorder, pre-processor and knowledge based discriminator including an inductive learning algorithm generating fuzzy decision trees. By their statistical evaluation feature vectors for training Kohonen networks are selected aplied for re-classification tests of clinical study data.