Continuous monitoring based on biosensors coupled with artificial intelligence
Résumé
Nowadays, continuous monitoring is a mandatory issue for almost all industries worldwide. For example, as a part of the quality control process, a large amount of samples is tested in food and beverage industries. Also, as a result of public concern about environmental pollution an accurate and faster monitoring is demanded by current legislations (Rodriguez-Mozaz, Marco et al. 2004). Traditional monitoring has been carried out by chromatographic techniques, which offer high accuracy, sensitivity and selectivity even in complex samples. But, even the outstanding characteristics of chromatographic techniques a faster, cheaper and reliable monitoring is highly desirable for applications where chromatographic techniques are not well suited (i.e. field monitoring, testing large amount of samples) (Rodriguez-Mozaz, Lopez de Alda et al. 2007) (Salvador, Adrian et al. 2007)
Some years ago, biosensors emerged as alternative tools for several monitoring applications involving the determination of a single analyte (i.e. toxin, biocide, metals) (Thevenot, Toth et al. 2001). However, as the number of analytes to monitor increase a multianalyte approach is now required. Usually, biosensors are designed to expose a high selectivity to a single analyte which makes difficult a multianalyte determination with a single device. Sensor arrays (i.e. sensors with different sensitivities and selectivities) have been proposed as an alternative to determinate several analytes with a single step (del Valle 2010). In the case of biosensors, arrays with different bioreceptors are integrated to achieve different selectivity and sensitivity to several analytes. But, since the higher order information provided by the biosensor array cannot be processed with the classical univariate approach a robust mathematical model is needed (Hierlemann and Gutierrez-Osuna 2008). In this sense, artificial neural networks have been successfully applied as multivariate calibration tool for such high order data to achieve a multianalyte determination(Bachmann, Leca et al. 2000) (Covaci, Sassolas et al. 2012) (Alonso, Istamboulie et al. 2012). In addition, artificial neural networks are able to model the nonlinearities found in sensor array providing a wider range of operation, even if the single devices do not provide conclusive information by themselves. For faster training, some authors have incorporated automatic flow injection systems to handle liquid samples(Gutes, Cespedes et al. 2007). The incorporation of flow systems, biosensor array and artificial neural networks to provide information has lead to new dedicated devices for automatic monitoring of pollutants (Valdes-Ramirez, Gutierrez et al. 2009, Crew, Lonsdale et al. 2011). These devices can be potentially used as monitoring tools in field applications providing continuous information of the analyzed samples.
This chapter attention is focus on the use of biosensor array for pollutant detection, emphasizing the use of artificial neural network as a multivariate calibration tool for multianalyte determination. In the first part, a brief description of biosensors as a single analytical tool is presented. Then, a complete review of monitoring of pollutants with biosensor arrays and the corresponding neural network modeling is covered. Finally, some fully integrated devices are presented and their potential applications as monitoring tools are discussed.
Domaines
Biochimie, Biologie Moléculaire
Origine : Publication financée par une institution
Loading...