Chemical imaging: I. Concepts and visions for electronic and bioelectronic noses

Abstract

Perfect ‘chemical imaging’ aims at the time- and spatially-resolved recording of many chemical species. Comparison of results from ‘chemical imaging’ with calibration data may also be trained towards an identification of odor impressions, environmental or medical conditions (such as toxicity), process control parameters etc. This ‘chemical imaging’ can be approached by either using the well-established techniques of analytical chemistry or by using a large number of calibrated sensors and sensor systems. The latter are sometimes denoted ‘electronic noses’, provide an electronic approach to artificial olfaction and are considered in this paper. They offer a variety of principal advantages including the fact that calibration efforts and sizes can be minimized systematically for specific applications by fine-tuning individual components of the sensor system. The paper describes a systematic to design such sensor systems. In the traditional application of chemical sensors the output of an individual chemical sensor is recorded as one ‘feature’. The first aim towards perfect ‘chemical imaging’ is to determine a large number of independent features, which span a large ‘hyperspace of chemical features’. The second aim is then to extract information from this hyperspace by optimizing a feature extraction procedure towards four application-specific goals. (a) The first goal concerns to record certain chemical species quantitatively and hence aims at perfect ‘chemical imaging’ as defined above. (b) Alternative goals concern to record odor impressions, (c) environmental or medical conditions, (d) and process control parameters. Different kinds of calibration are wanted to extract the wanted information from the data represented in the hyperspace of chemical sensor features. Hence, four different strategies are required to compare the features monitored by the chemical sensor systems with independent calibration standards from (a) instruments in analytical chemistry, (b) human odor panels, (c) (micro-)biological or medical tests, (d) and process parameter measurements. This adjustment of measured sensor features to calibration standards determines a specific type of feature extraction and pattern recognition for a specific application. This pattern recognition of experimentally recorded features is of key importance not only for these ‘electronic’ noses but occurs in the same way in all real ‘biological’ noses. Hence, formal analogies between the technical and biological world of noses are obvious. It is therefore expected, that our current studies on chemical sensor systems will also lead to a deeper understanding of signal processing in biological sensor systems and vice versa. Expected synergies of comparative studies concern in particular the molecular scale understanding of (a) the elementary processes of chemical sensing, (b) human odor perception, and (c) interactions between the environment and biological organisms. In this context, biolectronics becomes an increasingly important discipline. By taking advantage of characteristic similarities and differences of components in technical and biological systems, high-performance hybrid systems will be developped in the future.