Emabalajes que evitan contaminacion (1).pdf

of 7
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Available online at Sensors and Actuators B 128 (2008) 435–441 Self-organizing algorithm for classification of packaged fresh vegetable potentially contaminated with foodborne pathogens Ubonrat Siripatrawan ∗ Department of Food Technology, Faculty of Science, Chulalongkorn University, Bangkok, Thailand Received 26 March 2007; received in revised form 25 June 2007; accepted 27 June 2007 Available online 1 July 2007 Abstract A rapid method for identification of foodborne patho
   Available online at Sensors and Actuators B 128 (2008) 435–441 Self-organizing algorithm for classification of packaged fresh vegetablepotentially contaminated with foodborne pathogens Ubonrat Siripatrawan ∗  Department of Food Technology, Faculty of Science, Chulalongkorn University, Bangkok, Thailand  Received 26 March 2007; received in revised form 25 June 2007; accepted 27 June 2007Available online 1 July 2007 Abstract A rapid method for identification of foodborne pathogens contamination in packaged fresh vegetable using electronic sensor array and Kohonenself-organizingmap(SOM)algorithmwasdeveloped.  Escherichiacoli wasusedasthetargetmicroorganismbecauseitspresenceinfoodsindicatesfecal contamination, and the presence of pathogenic microorganisms.  E. coli  was grown in the packaged fresh vegetable. The electronic sensorswas used to monitor changes in the composition of the package headspace gas phase relating to the biochemical products of   E. coli  volatilemetabolites. SOM algorithm was then used to classify the data output from the electronic sensor array. The SOM algorithm created a map from ahigh dimensional input vector space onto a two-dimensional output lattice. When integrated with SOM algorithm, the electronic sensors proved tohave the ability to classify the packaged fresh vegetable potentially contaminated with pathogens.© 2007 Elsevier B.V. All rights reserved. Keywords:  Self-organizing map; Kohonen algorithm; Neural network; Foodborne pathogens; Contamination; Metal oxide sensors 1. Introduction Numerous outbreaks of foodborne diseases strengthen theneed for rapid and sensitive methods for detection of foodbornepathogens. Classical methods for the identification and clas-sification of microorganisms are based on their biochemical,morphological serological and toxigenic characteristics. Thesemethods usually require intact viable organisms and a seriesof tests requiring the incubation of the microorganisms [1,2].Early pathogens detection is important to implement diseasecontrol measures [3]. Recently, research has focused on devel- opment of rapid and accurate techniques to identify pathogensin food products [4–6]. Zhao et al. [7] developed a disposable electrochemical immunosensor for detection of   Vibrio para-haemolyticus  (VP) based on the screen-printed electrode (SPE)coated with agarose/Nano-Au membrane and horseradish per-oxidase (HRP) labeled VP antibody (HRP-anti-VP). Wu et al.[8]appliedQCMsysteminthedetectionofPCR-amplifiedDNAfrom real samples of   Escherichia coli  O157:H7. The piezoelec-tric biosensor detected the presence of   E. coli  O157:H7 when ∗ Fax: +66 2 2544314.  E-mail address: the DNA strand was complementary to the immobilized probeswith synthetic oligonucleotides.Microorganismscanbecharacterizedbyidentificationofspe-cific metabolites generated by specific biochemical pathways.The selection of volatiles for use as incipient disease indicatorshas been reviewed in terms of the composite rate of pathogenicdestruction within food products [9–11]. This concept has been actualizedinelectronicsensorarrayor“electronicnose”[12,13].A considerable number of electronic sensor applications havebeenreported,includingclassificationofchangesinmilkresult-ing from a variety of heat treatments [14], evaluation of the off-odor in wine [15], quality measurement of smoked salmon [16] and detection of   Salmonella  in nutrient media [17]. Elec- tronic sensor technology is usually based on a hybrid sensorarray system with different selectivity and sensitivity, with theresultbeingapowerfulanalyticalinstrumentespeciallyforcom-plex food analyses. However, analysis of volatile compoundsusing electronic sensor array often generates large and com-plicated data set and difficult to interpret if used directly. Amathematical resolution of complex data is usually performedin far less time than it takes to conduct physical or chemicalexperiments [17–19].Linear methods which have been used to classify thedata include linear discriminant analysis (LDA) and princi- 0925-4005/$ – see front matter © 2007 Elsevier B.V. All rights reserved.doi:10.1016/j.snb.2007.06.030  436  U. Siripatrawan / Sensors and Actuators B 128 (2008) 435–441 pal component analysis (PCA). The LDA and PCA is a lineartransformation that is well suited for separating image/signaldata for different objects or class [20]. The main advantage of linear transforms is that they are easy to design and typi-cally have closed-form solutions. However, linear transformstypically extract information from only the second-order corre-lations in the data (covariance matrix) and ignore higher-ordercorrelations in the data. Many researchers have suggested thatmany signals in the real world are inherently non-symmetric[21]. A number of nonlinear transformation methods for patternrecognition exist. Artificial neural networks (ANNs) are amongthe most commonly used nonlinear techniques.The most important features of ANN are their learning andadaptationabilities.Accordingtotheirlearningstrategies,ANNcan be classified as supervised and unsupervised networks. Insupervised learning, each time the ANN is exposed to a train-ing input, the related class information is required as well.The multilayer perceptrons (MLP) neural network or the feedforward ANN has been the most popular. The term ‘unsuper-vised’ means that the knowledge of environment is not learnedfrom the specific input–output examples. Self-organizing map(SOM) is an unsupervised artificial neural network which isfrequently used for data partitioning and classification. SOMcan be used for grouping of complex sample data withoutany strict assumption and without any priori knowledge of the number of groups present [20]. Lin and Wang [22] com- pared SOM with various hierachical cluster analysis methods.The result shows that the performance of SOM in clusteringmessy data is better than that of the other hierachical clusteringmethods.The principle of SOM is characterized by the formation of a topographic map of the input patterns in which the spatiallocations of the neurons in the lattice are indicative of intrin-sic statistical features contained in the input patterns [23]. The SOMcanbeconsideredasagridwithpredefinednodes.Priortolearning, a large unit area that surrounds the winner is selectedas a neighborhood region. During learning, the pattern of fillingthe nodes is determined by the degree of similarity between thedata. If an input vector is presented to the SOM network, theweight vector in the network that is closest to the input vec-tor is selected as the best-matching (winner) node. The winingmapping node is defined as that with the smallest Euclidean dis-tance between the mapping node vector and the input vector[24–26].Althoughvariousrapidmethodsfordetectionofmicroorgan-isms have been developed, no research has used electronic nosecoupled with SOM to classify the contamination of pathogensdirectly from the packaged food products. Hence, this researchwas aimed to develop a method to identify  E. coli  contamina-tion in packaged fresh vegetable using electronic sensor arraycoupled with Kohonen neural network.  E. coli  is a commonmember of the normal flora of the large intestine. In this study,  E. coli  was used as the target microorganism in packaged alfalfasprouts because its presence in foods indicates fecal contamina-tion, and the presence of pathogenic microorganisms .  Alfalfasprouts were chosen as the product component because theNational Advisory Committee on Microbial Criteria for Foods(NACMCF) [27] identified sprouts as a special problem due to the potential for pathogen growth during production, whilethere is increasing demand for sprouts due to their popularity asa healthy food [28,29]. The electronic nose was used to monitor thevolatilesproducedby  E.coli .SOMalgorithmwasusedasanexperimental platform (in addition to the instrumental methods)to identify  E. coli  contamination. 2. Materials and methods 2.1. Preparation of inoculated vegetable Thealfalfaseeds(NaturalSproutCompany,Springfield,MO)weresoakedin20,000ppmofcalciumhypochloritepriortoger-mination as advised by the U.S. Food and Drug Administration[28] and NACMCF [27]. Alfalfa sprouts were grown in a labo- ratory environment at 20 ◦ C and 65% RH with indirect sunlightand away from any possible contaminations. The sprouts wereharvested after 5 days (fully grown) when length is ∼ 3.8–4cm.The sprouts were washed and drained several times before use.Thenonpathogenicstrain  E.coli ATCC25922obtainedfromthe American Type Culture Collection (ATCC, Rockville, MD)was cultured in tryptic soy broth and incubated at 37 ◦ C for8h in a gyrotory shaker and centrifuged. Broth was pouredfrom the culture and the sedimented pellet was resuspendedin sterile Butterfield’s phosphate buffer which was used as adipping suspension. Preliminary experiments were conductedto determine the population of   E. coli  necessary in the dip-pingsuspensiontoresultinaninitialpopulationof  ∼ 10 5 CFU/gon sprouts. Preliminary studies also showed that the electronicnose was able to detect volatiles produced by  E. coli  whenthe number of   E. coli  was higher than 10 5 CFU/g. The sproutswere then placed in screened baskets, and submerged in thesuspension containing  E. coli  for 3min. The uninoculated con-trol was similarly treated except sterile phosphate buffer wasused in place of the inoculum. Fifty grams of sprouts werethen packed into commercial 1.5-mil, 15cm × 8cm linear low-density polyethylene (LDPE) bags and heat-sealed. The totalvolumeofthesproutswas ∼ 200mlwhichoccupiedabouthalfof thetotalbagvolume.Thesampleswereincubatedat10 ◦ Cfor1–3 days. 2.2. Microbiological analysis The microbial cell count was determined on the date of inoc-ulation and periodically throughout storage at days 1–3. Serialdilutions were prepared from the stock suspension, and Petriplates were inoculated with those dilutions expected to givecountable colonies. Inocula consisting of each of a dilution seri-als were deposited on prepared plates in duplicate using 3MPetrifilm Aerobic Count Plates (3M, St. Paul, MN) for deter-miningaerobicbacteriaand3MPetrifilm  E.coli  /ColiformCountPlates (3M, St. Paul, MN) containing Violet Red Bile nutrientagar as an indicator of glucuronidase activity for  E. coli . Allplateswereincubatedat37 ◦ Cfor48h.Platecountsarerecordedas colony forming units (CFU/g).  U. Siripatrawan / Sensors and Actuators B 128 (2008) 435–441  437 2.3. Electronic nose analysis A total of 120 samples, including non-inoculated alfalfasprouts (control) and alfalfa sprouts inoculated with  E. coli  attime zero, were incubated for 1–3 days prior to analysis.An electronic nose (Fox 3000, Alpha M.O.S., Hillsborough,NJ) was used for monitoring changes in volatiles produced by  E. coli  growing on the sprouts. The volatile analysis systemcombines a measurement chamber for generating the volatilecompounds and a detection system made up of 12 metal oxidesensors(SYLG,SYG,SYAA,SYGH,SYGCTI,SYGCT,T301,P101, P102, P401, T702, and PA2). This instrument was linkedto an auto-sampler capable of analyzing a total of 64 samples.Samples were placed in the HS100 auto-sampler in arbitraryorder. Five millilitre was collected from the headspace of pack-aged alfalfa sprouts and injected into the electronic sensor. Thetemperature of the injection syringe was 40 ◦ C. The delay timebetween two injections was 300s. Each injection was repeated,with separate samples. The electronic signals from the sensorswere digitized and then transferred to the control computer. 2.4. Self-organizing neural network  Dataweremadeupof120samplesfrom8subgroups(sprouts(SP) and sprouts inoculated with  E. coli  (EC) in LDPE bagson the first day of inoculation and incubated at 10 ◦ C for 1–3days). Each subgroup had 15 replicate samples collected fromseveral cultivations. Each sample was analyzed using 12 metaloxide sensors. The sensor responses of all 120 samples werearrangedina120 × 12matrix.Dataclassificationwasperformedusing SOM algorithm. All calculations were carried out usingMATLAB5.2routineswrittenbytheauthors,andmakinguseof the toolbox provided by Mathworks (Mathworks, Inc., Natick,MA). 3. Results and discussion 3.1. Microbial cell counts The number of aerobic bacteria and  E. coli  on the alfalfaseeds was determined. The number of aerobic bacteria was ∼ 10 1 –10 2 CFU/g,whileno  E.coli werefound.Thealfalfaseedsweresoakedin20,000ppmofcalciumhypochloritepriortoger-mination as advised by NACMCF [27]. This treatment has the potential to substantially reduce microbial contamination whichcanbepassedontothegrowingsprouts.Gilletal.[30]suggested that chemical disinfection can reduce the human risk for diseaseposedbycontaminatedseedsprouts.Thenumberofaerobicbac-teria in alfalfa seeds increased from ∼ 10 1 –10 2 to ∼ 10 7 CFU/gwhen the alfalfa sprouts were fully-grown. The conditions dur-ing sprouting (e.g. time, temperature, water activity, pH, andnutrient level) may have promoted the growth of microflora[28,30]  ,  without affecting the smell, taste or appearance of thesprouts. Thus, the risk of foodborne disease associated withsprouts increases during sprouting [27].Thecellcountsofaerobicbacteriaand  E.coli onfully-grownsprouts with and without  E. coli  inoculation are shown in Fig. 1. Fig. 1. Growth of total aerobic bacteria and  E. coli  of uninoculated vegetable(SP) and vegetable inoculated with  E. coli  (EC). All samples had a high number of total aerobic bacteria. How-ever,  E. coli  was not found in the control samples. The numbersof   E.coli intheinoculatedsamplesincreasedfrom ∼ 10 5 CFU/gon the first day of inoculation to ∼ 10 7 CFU/g after 3 days incu-bation. 3.2. Electronic sensor array Each sensor element changes in resistance ( Γ  max ) whenexposedtovolatilecompounds.Theinformationfromelectronicsensor array analysis was extracted from the series of sensorresistances.Inordertoproduceconsistentdatafortheclassifica-tion, the sensor response was presented with a volatile chemicalrelative to the base resistance in air, which is the maximumchange in the sensor’s electrical resistance divided by the initialresistance, as followsRelativeresistancechange = Γ  max − Γ  0 Γ  0 (1)where  Γ  max  is the maximum change in the sensor’s electricalresistance and  Γ  0  is the initial baseline resistance of the sen-sor. The relative resistance change was used for data evaluationbecauseitgivesthemoststableresult,andismorerobustagainstsensor baseline variation.The data matrix comprised 120 samples from 8 subgroups(SP-D0, SP-D1, SP-D2, SP-D3, EC-D0, EC-D1, EC-D2, andEC-D3) as analyzed using the 12 sensors (SYLG, SYG, SYAA,SYGH, SYGCTI, SYGCT, T301, P101, P102, P401, T702, andPA2). Fig. 2 shows the average responses of all samples in 8subgroups to the 12 metal oxide sensors. In Fig. 2, the sensi- tivities of all samples are compared. These values express theaverage sensor responses of each sensor in the range of mea-surement. Since the sensor outputs from the 12 different sensorsare not homogeneous, a direct comparison of sensitivities is notadequatetointerprettheinformationfromthesamples.Thesen-sor responses from an array of nonspecific metal oxide sensorsare generally insufficient to discriminate between a series of samples.  438  U. Siripatrawan / Sensors and Actuators B 128 (2008) 435–441 Fig. 2. Electronic sensor responses form the headspace of packaged sampleslabeledusingthefollowingscheme:SP-D0,SP-D1,SP-D2andSP-D3arepack-aged vegetable kept at 10 ◦ C for 0–3 days, respectively; EC-D0, EC-D1, EC-D2and EC-D3 are packaged vegetable inoculated with  E. coli  on the first day of inoculation and after stored at 10 ◦ C for 1–3 days, respectively. In this study, unsupervised SOM was used for analysis of electronicsensorarraydata.Although,thesupervisedMLPneu-ral network has been the most popular to model the complexdata, it requires prior training pairs (input vectors and corre-sponding target vectors) to make training possible. Therefore,theMLPmaybeunabletoprovideareal-timeresponsetodetectthe contaminated samples. Self-organizing map is an unsuper-visedneuralnetworkwhichdoesnotneedanyclassinformationfor learning, but acquires that knowledge by itself during thetraining phase through cluster formation. For SOM, the onlyinput is needed to construct an output. The SOM algorithm cre-ates a mapping from a high dimensional input vector space ontoatwo-dimensionaloutputlattice.TheSOMnetworkisbasicallycomposed of a single, two-dimensional layer of neurons whichhelps provide a visual presentation of data. 3.3. Self-organizing neural network  The unsupervised SOM was used to classify electronicvoltametric response outputs by replacement of the actualdata points using topographic map reference vectors. A two-dimensional Kohonen output layer was used to help provide avisual presentation. According to Lee et al. [31], selecting the appropriate number of output nodes is quite difficult and thisis usually experiment-dependent. There is no consensus amongresearchers about the subject. To obtain good mapping results,the number of output nodes in the Kohonen neural network should be at least 10–20% of training vectors. However, usingtoo few output nodes may cause the congestion of input vectorsover an output node, which may make it difficult to distinguishthe characteristics of the output space.AKohonennetworkconsistingof5 × 5nodeswasemployedfor classification of the 8 subgroups from the input data matrix(12 × 120). The predefined neuron number (grid size) in theKohonenoutputlayerwaschosenbecauseitwassufficienttodis-tinguish different sample groups. In SOM process, the mappingnodes are first initialized with random numbers. The SOM isinitialized by assigning small random values to all of the weightvector elements. The algorithm responsible for the formationof the SOM proceeds first by initializing the synaptic weightin the network. Once the network has been properly initialized,there are three essential processes involved in the applicationof the algorithm including sampling, similarity matching, andupdating [25,32,33].For the sampling process, a sample  x   from the input spacewas drawn with a certain probability. Let  x   denote an input vec-tor selected randomly from the input data space and  m  denotesthe dimension of the input space. The vector  x   represents theactivation pattern that is applied to the lattice. x = [ x 1 ,x 2 ,...,x m ] T (2)In the similarity matching step, the best matching, winningneuron i (  x  )attimestep t  wasdeterminedbyusingtheminimum-distance Euclidean criterion: i ( x ) = argmin|| x ( t  ) − w j  ( t  )|| (3)where w j   denotesthesynapticweightvectorofneuron  j. Ifan m -dimensional input vector is presented to the SOM network, thentheweightvectorinthenetworkthatisclosesttotheinputvectoris selected as the best-matching node. The particular neuron  i which is the best matching is called the winning neuron for theinput vector  x  .For the updating process, the synaptic weight vectors of allneurons were adjusted using the update formula: w j  ( t  + 1) = w j  ( t  ) + η ( t  ) h j,i ( x ) ( x i ( t  ) − w j  ( t  )) (4) h j,i ( x )  = exp  − d  2 j,i 2 σ  2   (5)where  x  ( t  ) is the input to node  i  at time  t   and  w j  ( t  ) is the weightfrom input node  i  to output node  j  at time  t  .  η ( t  ) is the learningrate parameter,  h  j,i (  x  ) ( t  ) is the neighborhood function centeredaround the winning neuron  i (  x  ),  d  2 j,i  is the distance between thewinningneuron i andtheadjacentneuron  j ,and σ   isthewidthof thetopologicalneighborhood[32].Thewinningnodeisselected asthecenterofaneighborhood,inordertoreducetheEuclideandistance. At time  t  , the cell learns this input signal. During thenexttime t+ 1,ithasaninformationprocessingabilityof  w j  ( t  + 1), which is close to the input signal. The neighboring units thatsurround  i (  x  ) also learn the input vector  x  ( t  ) by following thesame equation [33].For training, the data vectors are first arranged in randomorder and then presented in this order to the neural network fortraining. In SOM, the neurons adaptively tend to learn the prop-erties of the underlying distribution of the space in which theyoperate. Additionally, they also tend to learn their places topo-logically. The training consists of finding the winning neuron,whichistheonewhosepatternhasthebestmatchandmodifyingthe winning node and its closest neighbors in the neuron mapby moving their associated feature vectors closer to the inputvectors [25,32].


Jul 23, 2017


Jul 23, 2017
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks