Characterization of the short-term oxygen sensor signal response

Recent advances in biosensor technology resulted in the development of biosensors that can be implanted in living tissue for prolonged periods of time to monitor concentration of oxygen, glucose, and other metabolites. The implanted sensors function by electrochemical consumption of the metabolite producing continuous current signal that is proportional to blood metabolite concentration. Despite these advances, precise correlation of the sensor signal to the metabolite concentration still remains to be a challenge. Tissue signals are often contaminated by local fluctuations in metabolite concentration and no efficient algorithm that allows for characterization and removal of the signal noise has been developed. Characterization of the different types of variations observed in the in vivo tissue signals would be of tremendous value not only for implanted sensors, but also to stem cell implants, b-cell islet constructs and for various other tissue engineered cellular devices. The objective of this project was to design an algorithm that correlates the in vivo short-term oxygen sensor signal response to the blood oxygen levels. In vivo oxygen signals were pre-acquired from sensors implanted in hamster and pig animal models. Using these signals, we first identified and classified various fluctuations observed within the oxygen signals. The signal variations were classified as biological or non-biological based on a review of the physiological ranges of oxygen fluctuation within tissue. Next, a filtering algorithm was designed to specifically remove the non-biological signal features from the oxygen signal to prevent the signal analysis from being distorted and contaminated by the these unwanted features. Finally, digital and statistical signal processing, and time series analysis methods were used to characterize the filtered oxygen signal. The power spectrums of the filtered oxygen signals were estimated, and interestingly, an oscillatory pattern was identified that supports a fundamental frequency of 10 +/- 5 mHz for arteriolar vasomotion within the tissue. Correlation analysis revealed periodic cycles of regional tissue oxygen perfusion ranging from 0.29 to 2.1 mHz. The signals were also analyzed using continuous wavelet transforms, and the short-term oxygen signals were determined to be both stationary and non-stationary. These results strengthen the hypothesis that the short-term oxygen signals are composites of many different types of biological variations which occur at different time intervals within the tissue. Significantly, probability distribution analysis showed that the oxygen signals collected from the same sensor array share the same non-Gaussian probability distribution properties. Using these characterizations, the effect of the local heterogeneous tissue environment can be removed from the short-term in vivo oxygen signals creating an oxygen signal representative of the global blood oxygen environment. Therefore, the design of an algorithm to characterize the tissue sensor signal variations is extremely valuable and will ultimately lead to design of highly accurate implantable metabolite tissue sensors.