Tailored xerogel-based sensor arrays and artificial neural networks yield improved O₂ detection accuracy and precision

The objective of this research is to develop arrays of tuned chemical sensors wherein each sensor element responds to a particular target analyte in a unique manner. By creating sol-gel-derived xerogels that are co-doped with two luminophores at a range of molar ratios, we can form suites of sensor elements that can exhibit a continuum of response profiles. We trained an artificial neural network (ANN) to "learn" to identify the optical outputs from these xerogel-based sensor arrays. By using the ANN in concert with our tailored sensor arrays we obtained a 5-10 fold improvement in accuracy and precision for quantifying O₂ in unknown samples. We also explored the response characteristics of these types of sensor elements after they had been contacted with rat plasma/blood. Contact with plasma/blood caused ∼15% of the luminophore molecules within the xerogels to become non-responsive to O ₂. This behavior is consistent with rat albumin blocking certain pore sub-populations within the mesoporous xerogel matrix thereby limiting O ₂ access to the luminophores.