Sensor’arraysfor liquid’sensing’ …w3.ualg.pt/~hgomes/html/classes/Biosensores/electronic_tongue/e...Sensor’arraysfor liquid’sensing’ (electronic’tongue’systems)’

Sensor arrays for liquid sensing (electronic tongue systems)

Henrique Leonel Gomes Universidade do Algarve, FCT, Campus de Gambelas,

8000 Faro, Portugal ([email protected])

Biossensores, Mestrado Integrado em Engª. Electrónica e Telecomunicações (MIEET-‐2009/00)

Introduc:on

Electronic tongue systems are mulOsensor devices dedicated to automaOc analysis of complicated composiOon samples and to the recogniOon of their characterisOc properOes. Many possible architectures of such devices were proposed: impediometric, potenOometric, voltammetric, as well as approaches embracing mass and opOcal-‐sensors. For the analysis of sensor array data, various paSern recogniOon systems were proposed.


Sources used for this lesson

•  A paper by Patrycja Ciosek and Wojciech Wróblewski , Analyst, 2007, 132, 963–

978.

•  A tutorial on Principal Components Analysis by Lindsay Smith. •  InformaOon and visualizaOon techniques for sensing and biosensing •  hSp://www.icmc.usp.br/~paulovic/pexsensors/ •  The work of our reaserch group


Outline


•  Sensor arrays in electronic tongue systems. •  Sensor fusion: hybrid electronic tongues and ET/EN devices. •  The analysis of sensor array responses – paSern recogniOon

system. •  •  Examples (an electronic tongue to classify honey)

Learning outcomes •  Know the operaOng principle of a E-‐tongue or a E-‐nose. •  For a given a complex substance to analyze the student should be able to devise

the best array of sensors to be used (opOcal, impedance, potenOometric, etc.). •  Know how to use staOsOcal tools or paSern recogniOon systems to treat the

arrays of data provided by the sensors (example principal component analysis).

•  Design circuits to process informaOon from arrays of sensors.

Defini:ons The electronic nose (EN) and electronic tongue (ET) (synonyms: arOficial nose,

mechanical nose, odour sensor, taste sensor, taste system, taste chip) are systems, whose construcOon and principle of operaOon were inspired by the neurophysiology of the senses of smell and taste. They are dedicated to the automaOc analysis of complicated composiOon samples, to the recogniOon of their characterisOc properOes, and generally they are assigned to fast qualitaOve analysis. They consist of an array of sensors exhibiOng various selecOvity and paSern recogniOon systems that analyse the sensor responses.


Electronic tongue

The construcOon of and electronic tongue demands the fusion of knowledge from: sensory technologies: •  PaSern recogniOon methods •  ArOficial intelligence •  Chemometric tools. •  Circuit design


A wide variety of chemical sensors can be employed in the design of electronic tongues: Electrochemical, voltammetric, potenOometric, Impedance OpOcal or enzymaOc sensors

Types of electrodes used in e-‐tongues (Ion-‐selecOve electrodes)

The principle of operaOon of the ion-‐selecOve electrodes is based on the measurement of their potenOal changes against a reference electrode in zero-‐current condiOons. The potenOal of the ion-‐selecOve electrodes is a funcOon of the acOvity of ionic species in a sample soluOon and is formed in the ion-‐sensiOve membrane, where selecOve complexaOon (ion recogniOon) of the analyte molecules occurs.

The main disadvantages of potenOometric measurements are temperature dependence, the influence of soluOon change, and adsorpOon of soluOon components that affect the nature of charge transfer, but the effect of those factors can be minimized by the control of temperature.

Disadvantages:

The principle of operaOon


Descrip:on of the cross-‐sensi:vity (The average sensor slope S )

The mean sensiOvity for various species):

where Si is the sensor response slope in a soluOon of an individual ion, one of n ions in total.

The second coefficient, s, describes the average value of formal sensiOvity, the signal-‐to-‐noise raOo for each ion: where s is the standard deviaOon of the i response slope in soluOons of each ion.

The last parameter, F, describes the sensiOvity distribuOon of a sensor towards the invesOgated ions. It is called a ‘non-‐selecOvity factor’ and is calculated according to the formula:


Voltammetric and other electrochemical sensors for ET systems

Voltammetric measurements are performed when equilibrium is not reached, and the signal obtained is the current-‐potenOal relaOonship. The simplest measurement set-‐up employs three electrodes: reference, working and auxiliary electrodes. The potenOal of the reference electrode is assumed to be constant, and between the working and auxiliary electrodes the current flows. The electrolysis reacOon occurs on the working electrode and that process is responsible for current generaOon. The current is a funcOon of the rate of electrolysis, which in turn is governed by the transport of electroacOve species present in the sample (i.e. diffusion coefficients and concentraOons of electroacOve species).


Sensors for e-‐Tongues (Op:cal sensor arrays)

Species, which are difficult to detect electrochemically (e.g. are not charged and/or are not electroacOve) can omen be analyzed with the use of opOcal sensors. There are many possible modes of operaOon of opOcal sensors: the acquisiOon of: •  Fluorescence intensity. •  LifeOme signals. •  Absorbance. •  Reflectance.


Sensor fusion: hybrid electronic tongues and ET/EN devices

The fusion of various measurement techniques was recently proposed to improve the recogniOon capabiliOes

Taste impression = smell + taste + texture + color + sound + temperature

Therefore, the complete characterizaOon of a sample would demand not only chemical characterizaOon, but physical sensors should also be considered (pressure/tacOle sensors, acousOc sensors, temperature sensors).


The analysis of sensor array responses paIern recogni:on system

The percepOon of sOmuli performed by the receptors demands appropriate transporOng and processing by the neural system

Its task is to recognize the sOmuli and properly react towards them. In arOficial chemical senses, where sensors play the role of the receptors, those acts are performed by various numeric procedures realized by a computer. Those procedures form a so-‐called PaSern RecogniOon System (PARC System). Its aim is to recognize the invesOgated objects, to categorize between various types of them, and to classify objects to a given set, i.e. class of objects

PARC tools employ various mathemaOcal, staOsOcal, machine learning, and signal processing methods.


PaIern Recogni:on System In pracOce, this is realized in two steps: First, the data are preprocessed in order to: •  make it independent from units, •  remove redundant informaOon, •  enhance signal-‐to-‐noise raOo. Amer that, the model describing the relaOon between X and Y has to be created. (In most cases, PCA is used as a preprocessor for sensor array data).

It decomposes the data matrix into a new set of uncorrelated variables (Principal Components), by finding new direcOons in the paSern space, so that they can explain the maximum amount of variance within the data set These new variables may be ploSed on a PCA plot or used as inputs for more complex classifiers, e.g. neural networks.


What is PCA analysis ?

It is a way of idenOfying paSerns in data, and expressing the data in such a way as to highlight their similariOes and differences. Since paSerns in data can be hard to find in data of high dimension, where the luxury of graphical representaOon is not available, PCA is a powerful tool for analysing data.

The other main advantage of PCA is that once you have found these paSerns in the data, and you compress the data, ie. by reducing the number of dimensions, without much loss of informaOon. This technique used in image compression, as we will see in a later secOon.

Will take you through the steps you needed to perform a Principal Components Analysis on a set of data.


PCA analysis

Method Step 1: Get some data Step 2: Subtract the mean For PCA to work properly, you have to subtract the mean from each of the data dimensions. The mean subtracted is the average across each dimension. So, all the values have (the mean of the values of all the data points) subtracted, and all the values have subtracted from them. This produces a data set whose mean is zero.


PCA analysis

PCA example data, original data on the lem, data with the means subtracted on the right, and a plot of the data


PCA analysis

Step 3: Calculate the covariance matrix

This is done in exactly the same way as was discussed in secOon 2.1.4. Since the data is 2 dimensional, the covariance matrix will be . There are no surprises here, so I will just give you the result:

Step 4: Calculate the eigenvectors and eigenvalues of the covariance matrix


PCA analysis

Step 4: Calculate the eigenvectors and eigenvalues of the covariance matrix

It is important to noOce that these eigenvectors are both unit eigenvectors ie. their lengths are both 1. This is very important for PCA, but luckily, most maths packages, when asked for eigenvectors, will give you unit eigenvectors.


PCA analysis

So what do they mean? If you look at the plot of the data in Figure then you can see how the data has quite a strong paSern. As expected from the covariance matrix, they two variables do indeed increase together. On top of the data I have ploSed both the eigenvectors as well. They appear as diagonal doSed lines on the plot. As stated in the eigenvector secOon, they are perpendicular to each other. But, more importantly, they provide us with informaOon about the paSerns in the data. See how one of the eigenvectors goes through the middle of the points, like drawing a line of best fit? That eigenvector is showing us how these two data sets are related along that line. The second eigenvector gives us the other, less important, paSern in the data, that all the points follow the main line, but are off to the side of the main line by some amount.

A plot of the normalised data (mean subtracted) with the eigenvectors of the covariance matrix overlayed on top.


PCA analysis

Step 5: Deriving the new data set

This the final step in PCA, and is also the easiest. Once we have chosen the components (eigenvectors) that we wish to keep in our data and formed a feature vector, we simply take the transpose of the vector and mulOply it on the lem of the original data set, transposed.

Where Row Feature vector is the matrix with the eigenvectors in the columns transposed so that the eigenvectors are now in the rows, with the most significant eigenvector at the top, and Row Data Adjust is the mean-‐adjusted data transposed.


PaIern Recogni:on System ParOal Least Squares-‐Discriminant Analysis (PLS-‐DA)

In contrast to PCA, ParOal Least Squares-‐Discriminant Analysis (PLS-‐DA) is a supervised method which models the relaOonship between two matrices, i.e. X and Y. PLS-‐DA determines a set of latent variables, corresponding to principal components in PCA, but explaining as much of the covariance as possible between the two matrices (PLS-‐DA scores)

It is a generalizaOon of mulOple linear regression, it can analyze more noisy and uncompleted data and it is able to manage with mulO-‐colinearity problem, which omen occurs in sensor array measurements.


Informa:on and visualiza:on techniques for sensing and biosensing

hSp://www.icmc.usp.br/~paulovic/pexsensors/

InformaOon and visualizaOon techniques for sensing and biosensing


Electronic tongue at UAlg (classificaOon of honey)


ClassificaOon of honey (eletcronic tongue at UAlg)

PCA analysis of different types of honey, Lar (laranjeira), Med (Medronho), Gir (Girassol) e Ros (Rosmaninho)


Applica:ons of Electronic toungues

Source: Patrycja Ciosek and Wojciech Wróblewski, Analyst, 2007, 132, 963–978


Commercial systems

Source: Patrycja Ciosek and Wojciech Wróblewski, Analyst, 2007, 132, 963–978


Interface electronic circuitry for an electronic tongue

They are strongly dependent on the type of the sensor array used They relay strongly on Integrated soluOons ADC and DACs Low-‐power and noise are important. InteresOng project for instrumentaOon course!


Summary •  To construct an electronic tongue you need an array of sensors ( what

they really measure is not crucial and fundamental to understand). •  For instance wine and coffee are extremely complex substances, It is

difficult to get insight into the physical interacOons with the electrodes or with the sensors.

•  The sensors must be selected according to the liquid to be measured.

(you may need the helps of a chemist or a physicist to selected the most appropriate sensors.

•  Once the sensors are selected you need a readout circuit that process the data from the sensor array.

•  The data is treated by paSern recogniOon systems. There are a few around (Principal component analysis is a omen used method)

Documents

Sensor’arraysfor liquid’sensing’ …w3.ualg.pt/~hgomes/html/classes/Biosensores/electronic_tongue/e...Sensor’arraysfor liquid’sensing’ (electronic’tongue’systems)’