Total Polyphenols in Green Tea Samples by FT-NIR Spectroscopy

Transcription

1 Fresh Produce 009 Global Scece Books Total Polypheols Gree Tea Samples by FT-NIR Spectroscopy Vadakkepulppara Ramachadra Nar Sja * Har Nwas Mshra ** Agrcultural & Food Egeerg Departmet, Ida Isttute of Techology, Kharagpur, West Begal-71 30, Ida Correspodg authors: * sjavr@redffmal.com ** hm@agfe.tkgp.eret. ABSTRACT The feasblty of measurg total polypheols cotet stat gree tea powder ad gree tea graules was vestgated by Fourer Trasform Near-Ifrared (FT-NIR) spectroscopy. The spectra were measured dffused reflectace mode by keepg 8-10 g samples a small sample bottle. A partal least square regresso model was developed wth vector ormalzato as the pre-processg method the NIR rego ( cm -1 or m). The developed model was valdated usg a cross valdato techque. FT-NIR spectroscopy wth chemometrcs, usg PLS vector ormalzato as the pre-processg method could predct the total polypheols cotet tea samples terms of gallc acd accurately up to a R value of ad a stadard error of cross valdato (RMSECV) value of 1.45 wth 4 factors the predcto model. The developed model was appled to predct total polypheols gree tea samples wth m. The developed procedure was further valdated wth fresh samples whch were ot used for calbrato ad compared wth spectroscopc method of ploypheol determato. The overall results demostrate that NIR spectroscopy wth multvarate calbrato could be successfully appled as a rapd method ot oly to detfy tea varetes but also to determe total polypheols cotet gree tea samples. Keywords: chemometrc aalyss, Fourer trasform ear frared spectroscopy, stat tea, PLS regresso model, spectral preprocessg INTRODUCTION Tea (Camella sess) s the secod most cosumed beverage the world after water (L ad He 008). Wth the creasg cosumpto of tea, ts qualty cotrol has becomes more ad more mportat owadays, for example, may atoal ad teratoal authortes are settg crtera for qualty factors lke total polypheols ad caffee (Ary 1999; Che et al. 006). I geeral, caffee ad total polypheols are aalyzed as the mportat qualty factors for tea leaves. These costtuets (polypheols ad caffee) are maly resposble for the characterstc astrget ad btter taste of tea brews (Zhag et al. 199). Addtoally, most commercally tea leaves have may varetes o the market, whch dffer ot oly from a botacal stadpot but also terms of qualty. These dffereces are recogzed commercally ad also apprecated by cosumers. I the past few years, may dfferet methods of aalyss have bee employed to detfy tea varetes ad to determe the chemcal composto of tea. Some approaches were appled to detfy tea varetes usg moder techques lke hgh-performace lqud chromatography (HPLC), gas chromatography (GC), plasma atomc emsso spectrometry, FTIR-ATRI, Rama Spectroscopy, CE- MS/MS, LCMS, etc., (Togar et al. 1995; Hore ad Kohata 000; Zuo et al. 00; Satose et al. 005; Sawalha et al. 009). Also some approaches were appled to quattatvely aalyse the chemcal composto of tea leaves, such as HPLC (Zuo et al. 00), capllary electrophoress (Hore et al. 1997), colormetrc measuremets (Che et al. 009) ad ttrato wth potassum permagaate (ISO 1994). However, all of the methods metoed above are tmecosumg. Near frared (NIR) spectroscopy has proved to be a powerful aalytcal tool used the agrcultural, utrtoal, petrochemcal, textle ad pharmaceutcal dustres (McGloe et al. 00; Dez 004; Woodcock et al. 008; Ferádez-Ibañez et al. 009). Sce the 1990s, attempts have bee made to smultaeously predct water, alkalod ad pheolc substace cotet tea leaves usg NIR spectroscopy (Hall et al. 1988; Schulz et al. 1999). Studes o the applcato of NIR spectroscopy to quattatve aalyss of total atoxdat capacty gree tea was also reported (Lupaert et al. 003; Zhag et al. 004). Although they provded some better results (more sestve ad rapd) for tea usg NIR spectroscopy, they dd ot dscuss detals of the predcto models eve wthout usg depedet test samples to test the robustess of the models, such as Schulz et al. (1999). Gree tea cotas polypheols, whch clude flavaols, flavadols, flavoods, ad pheolc acds; these compouds may accout for up to 30% of the dry weght. Most of the gree tea polypheols are flavools, commoly kow as catechs. Tea polypheols are of great terest due to ther beefcal medcal propertes (Yag et al. 00). There s creasg evdece that polypheols substaces foud tea ca ehace geeral health. Recetly, may researches have suggested that atoxdats foud polypheolc substaces may play a mportat role the preveto of cardovascular dsease (Nakach et al. 000), chroc gastrts (Shbata et al. 000; Setawa et al. 001) ad some cacers (Fujk et al. 001; Ioue et al. 001; Ja et al. 004). Addtoally, polypheolc compouds are maly resposble for the characterstc astrget ad btter taste of tea brews. I recet years, may methods of aalyss have bee employed to determe total polypheols cotet tea, such as colormetrc measuremets ad the ttrato method wth potassum permagaate (ISO 1994). However, these methods are all tme-cosumg. NIR spectroscopy s a fast, accurate ad o-destructve techque that ca be employed as a replacemet of tme-cosumg chemcal methods. I all other chemcal ad aalytcal Receved: 1 Jauary, 009. Accepted: 1 May, 009. Orgal Research Paper

2 Fresh Produce 3 (1), Global Scece Books methods sample preparato s eeded ad t s destructve also. Tme take for dvdual measuremet s very log. I FT NIR oce the calbrato s over, t wll take oly secods to determe the amout of a partcular compoet the sample wthout ay sample preparato. The objectve of preset study was to develop a rapd method for the quattatve estmato of total polypheols cotet stat gree tea powder ad gree tea graules usg FT-NIR spectroscopy. I commercal producto, we ca use ths method to aalyze the amout of total polypheols the fal product ad also to check the same durg packagg ad storage as a very rapd ad o-destructve method. The motorg ad determato of polypheols cocetrato tea products s essetal for maufacturers because most of the beefcal effects of tea are maly due to the polypheols. MATERIALS AND METHODS Sample preparato The fresh tea leaves for preparato of tea samples were plucked from the tea gardes of the Ida Isttute of Techology, Kharagpur, Ida. Fresh tea leaves were steamed (1 kg/cm for 1- m) mmedately after pluckg to arrest the fermetato process (oxdato reacto) ad the groud a laboratory grder. From the paste thus obtaed, a porto of the juce (40-45%) was extracted by meas of hydraulc pressg a extracto ut developed at the Agrcultural & Food Egeerg Departmet of IIT Kharagpur, Ida (Sja ad Mshra 009). The juce wth a total sold cotet of 6-9% was used to produce stat tea powder samples by freeze dryg ad a pressed leaf resdue wth a mosture cotet of 63-65% wet bass (wb) was subjected to hot ar dryg (temperature: C, thckess: 3-5 mm, ar velocty: m/s) a recrculatory covectve ar drer (Bose Istrumets Pvt. Ltd., Kolkata, Ida) to produce dfferet gree tea graules samples. Sce dfferet dryg codtos were employed, samples wth dfferet amouts of total polypheols thus resulted. A FT-NIR MPA TM spectrometer (Bruker Optcs, Germay) combed wth Opus 5.5 software was used for aalyss of tea samples by geeratg a uque spectrum for each sample. Ths spectrometer wth a tegrated Mchelso terferometer utlzed the Fourer-Trasform ad had dstct advatages compared to dspersve spectrometers. The Mchelso terferometer FTNIR has two basc advatages. Frst, the Fellgett multplex advatage derves from the smultaeous processg of the etre spectral rage durg a sgle sca. All frequeces the spectra are measured smultaeously Fourer trasform ear-frared (FTNIR) spectrometer for the etre tme. Ths s because a terferometer ca modulate at frequeces that are proportoal to the wave legth. The tme advatage s eve larger, sce t s drectly proportoal to spatal elemets examed. A complete spectrum ca be collected very rapdly, ad may scas ca be averaged the tme take for a sgle sca of a dspersve spectrometer. Secod, the Jacquot advatage allows for a large eergy throughput because t s possble to use a large aperture. For the same resoluto, the eergy throughput a FTNIR spectrometer ca be hgher tha a dspersve spectrometer, where t s restrcted by the slts. I combato wth the multplex advatage, ths leads to oe of the most mportat features of a FTNIR spectrometer: the ablty to acheve the same sgal-to-ose rato as a dspersve strumet a much shorter tme. Also, the trsc wavelegth scale a FTNIR spectrometer provdes wavelegth repeatablty better tha oe part a mllo (Coes advatage). The wave umber calbrato of terferometers s also much more precse (Bra 001; Wllam ad Norrs 001). Chemometrcs: multvarate aalyss Multvarate aalyss was used for quattatve ad qualtatve aalyss. A Partal Least Square (PLS) algorthm, whch was prove to be effectve may quattatve applcatos (determato of total solds, ph, caffee tea (Che et al. 006)), was used the preset study. The OPUS 5.5 software was used for PLS aalyss. These methods wth orgal ad vector ormalsed spectra were used to develop calbrato models. The performace of the fal PLS model was evaluated terms of root mea square error of cross-valdato (RMSECV) for cross valdato ad root mea square error of predcto (RMSEP) durg test valdato, ad the coeffcet of determato (R ). For RMSECV, a leave-oesample-out cross-valdato was performed: the spectrum of oe sample of the trag set was deleted from ths set ad a PLS model was bult wth the remag spectra of the trag set. The left-out sample was predcted wth ths model ad the procedure was repeated by leavg out each of the samples of the trag set. I test valdato a set of samples were detfed as test data ad wth the remag data set a calbrato model was developed frst ad the the test spectra was used for the valdato of the developed model (Sja ad Mshra 009). The umber of PLS vectors used s defed the OPUS software by the sze of the rak. The optmum PLS rak ca be calculated oly f the umber of calbrato spectra s suffcetly hgh (e.g. oe compoet ad 0 calbrato spectra). The PLS regresso has the advatage that the PLS factors are arraged correct sequece, accordg to ther relevace to predct the compoet values. The frst factor explas the most drastc chages of the spectrum. The resdual (Res) s the dfferece betwee the true ad ftted value. Thus the sum of squared errors (SEE) s the quadratc summato of these values (Eq. 1). SSE RMSEE R [Re s ]...(1) The root mea square error of estmato (RMSEE) s calculated from ths sum, wth beg the umber of samples ad r the rak (Eq. ). 1 SSE...() r 1 The determato coeffcet, R (Eq. 3) gves the percetage of varace preset the true compoet values, whch s reproduced the regresso. R approaches 100% as the ftted cocetrato values approach the true values. SSE ( y y 1 ) m RMSECV RMSEP 1 1 ( y ( y (3) where y m s the mea of the referece results for all samples. R ca be egatve ( some cases) for low raks, whe the resdual are larger tha the varace the true values (y ). I case of cross valdato the RMSECV s calculated usg Eq. 4. y ) yˆ )...(4) where s the umber of samples the trag set, y the referece measuremet result for the sample ad y s the estmated result for sample whe the model s costructed wth the sample removed. The umber of PLS factors cluded the model s chose accordg to the lowest RMSECV. Ths procedure s repeated for each of the pre-processed spectra. For the test set, the RMSEP s calculated as follows (Eq. 5)....(5) where, s the umber of samples the test set, y the referece measuremet result for test set sample ad ŷ s the estmated result of the model for test sample. Preparato of calbrato ad valdato models Spectra of 30 dfferet gree tea samples (wth dfferet polypheol cotets, 14-9%) prepared were used for model developmet. The samples were prepared wth dfferet raw materals (dfferet 77

3 Total polypheols by FT-NIR spectroscopy. Sja ad Mshra varety ad dfferet pluckg levels) ad varyg processg codtos order to have dfferet polypheol cotet. The NIR spectra were collected the reflectace mode usg the FT-NIR MPA TM spectrometer. The rage of spectra was from 1,000 to 4000 cm 1. The stadard sample accessory holder was used for performg the tea spectra collecto. For each tea sample, 8-10 g of dry tea leaves was flled to the sample cup the stadard procedure depedg upo the bulk desty of materals. The correspodg amout of dry tea powders was desely packed to the sample cup ad the compressed by closg t. For each sample, three spectra were recorded at three dfferet pots by rotatg the sample bottle by 10. The average of the three spectra, whch were collected from the same tea sample, was used the ext aalyss. The temperature was kept aroud 5C ad the humdty was kept at a steady level the laboratory. The amout of total polypheols each sample was determed by the stadard method usg Fol Cocalteau reaget whch s descrbed below. Cross valdato method was used ths model. Estmato of total polypheols by spectrophotometrc method Fve grams of gree tea sample was groud well a mortar ad pestle wth 100 ml of 95% ethaol ad the volume made up to 50 ml. Two ml of ths extract was dluted to 100 ml wth dstlled water. I a bolg test tube ml of the dluted extract, 4 ml of Foll Cocalteau (FC) reaget ad ml of sodum carboate solutos were added, dluted to 1 ml wth dstlled water ad shake well. Ths mxture was allowed to stad for 30 m at room temperature. After that the blue colour developed the sample at 760 m was read by a spectrophotometer at room temperature. Pheols react wth phosphomolybdc acd FC reaget alkale medum ad produce a blue colored complex. A stadard curve was prepared wth gallc acd for total polypheol estmato the cocetrato rage betwee 0 ad 30% (Sadasvam ad Mackam 1996). Valdato usg fresh samples Polypheol cotet of fresh gree tea sample (ay sample wth polypheol cotet dfferet from the set used for calbrato, but wth the rage used for calbrato) was determed usg the spectrophotometrc method order to verfy the accuracy of the developed method. Trplcate measuremets were take for the same sample ad results were aalyzed statstcally. Oe-way ANOVA was carred out for the result obtaed by both the methods ad p value was calculated, from whch t s clear whether there s sgfcat dfferece betwee the results obtaed by two methods. Trplcate measuremets were take for the same sample ad the average value was used for comparso. RESULTS AND DISCUSSION Fg. 1 shows the FT-NIR spectra of tea samples used for developg the calbrato model. These true peaks were selected after smoothg the spectrum to avod terferece due to ose. The most tesve bad the spectrum belogs to the vbrato of the secod overtoe of the carboyl group (585 cm -1 ), followed by the -CH (7137 cm -1 ), the -CH (547 cm -1 ) ad the -CH 3 overtoe (5808 cm -1 ). The vbrato of the C=O, -CH ad -CH are caused by gredets such as polypheols, alkalods, prote, volatle as well as o-volatle acds ad by some aroma compouds (Paradkar ad Irudayaraj 00). The NIR rego cotas bads that ofte overlap, makg t dffcult to extract spectral parameters of the dvdual bads. Chemometrcs have provded a way of overcomg these problems through emprcal models that relate the multple spectral testes from may calbrato samples to kow aalytes these samples. As the spectra show smlar basc FTNIR spectral patters, mathematcal trasformatos were requred to use the FTNIR data for quattatve aalyss. Despte the lack of dstct peaks, t has bee show the PLS ca extract relevat formato for quattatve determatos (McShae ad Cote 1998). Fg. 1 FT-NIR absorpto spectra of calbrato data set. Dfferet colours correspod to the dfferet samples (30 samples used for ths study). Coeffcet of determato R RMSECV PLS factor Fg. RMSECV ad r plotted as a fucto of PLS factors. Predcted total polypheol cotet (%) y = x R = Measured total polypheol cotet (%) Fg. 3 Lear regresso plot of measured versus predcted cotet of total polypheols. I the applcato of PLS algorthm, t s geerally kow that spectral preprocessg methods ad the umber of PLS factors are crtcal parameters. Here ther effects o the results are dscussed. The optmum umber of factors s determed by the lowest RMSECV ad hghest value for r. Table 1 shows the values of r ad RMSECV for dfferet pre-processg techques used. From the table t s clear that compared wth others, the lowest RMSECV value equals 1.45 obtaed after the vector ormalzato spectral pre-processg ad also maxmum value for coeffcet of determato. Fg. shows values of RMSECV ad R cor RMSECV 78

4 Fresh Produce 3 (1), Global Scece Books Table 1 RMSEP ad R values correspodg to each PLS factor for determg caffee cotet wth dfferet spectral pre-processg methods. Pre-processg techque R (valdato) RMSECV R (calbrato) RMSEE PLS factor No pre-processg Vector ormalzato st dervatve Straght le subtracto d dervatve Table Comparso of results obtaed by FTNIR method ad spectroscopc method for total polypheol determato. Summary Groups Cout Sum Average Varace Spectrometrc method FT-NIR method ANOVA Source of Varato SS df MS F P-value F crt Betwee Groups NS Wth Groups Total NS o sgfcat respodg to each PLS factor for determg total polypheols wth vector ormalzato as the spectral pre-processg method. See from fgure t s clear that the maxmum value of R ad mmum RMSECV value s for PLS factor 4 vector ormalzato method. Ths model eeds four PLS factors. I ths applcato, vector ormalzato seems to perform better tha other pre-processes. The NIR predcted results for total polypheol cotet gree tea samples usg the calbrato model s preseted Fg. 3. The PLS-regresso method gave R values of for calbrato data set ad RMSECV value of 1.45 for cross valdato. The results of ths study clearly demostrated the capablty of FTNIR for ths applcato. The predctve abltes of the calbrato models were evaluated by the resdual predctve devato (RPD) also, whch was defed as the rato of the stadard devato of the referece data to the stadard error of predcted data for the populato tested (Pk et al. 1998). RPD value for the preset model s 4.5, whch s greater tha three ad cosdered to be desrable for predcto purposes as per the prevous refereces (Pk et al. 1998; Neuwoudt et al. 006). Aalyss of tea samples The result obtaed by the developed method was valdated wth fresh tea sample ad cross checked by stadard spectrophotometrc method. Total polypheol cotet of freshly prepared tea samples were measured usg the above two methods. Total polypheol cotet values for each sample was measured trplcate by the above methods ad the values obtaed were aalyzed statstcally. The results of the ANOVA (Table ) shows that F crtcal s greater tha F tabulated ad that the p value s , whch shows that there s o sgfcat dfferece betwee the values obtaed by the two methods eve at 5% level of sgfcace. Ths proved that FTNIR spectroscopy s a rapd ad effcet tool for detecto ad quatfcato of mosture cotet gree tea samples. CONCLUSIONS A rapd ad smple FT-NIR procedure to estmate the amout of total polypheols gree tea samples was developed usg a sgle calbrato model. The model was developed usg the spectral rego cm -1. The maxmum coeffcet of determato (R ) value of ad RMSECV value of 1.45 was obtaed for model developed for predcto of mosture cotet. The performace of developed method was cofrmed wth freshly prepared samples ad the values obtaed for total polypheol cotet of samples by ths method was foud to have o sgfcat dfferece wth the values obtaed by spectrophotometrc method. Tme take for dvdual measuremet s betwee secods. Ths method ca be adopted drectly by the dustres order to determe the total polypheolc cotet of gree tea samples at varous stages of processg to check the qualty ad also at the tme of packagg or durg storage wthout destroyg the sample. ACKNOWLEDGEMENT The frst author s thakful to All Ida Coucl for Techcal Educato (AICTE) New Delh, for the facal assstace receved the form of Natoal Doctoral Fellowshp to pursue her Doctoral programme. REFERENCES Ary A (1999) Sub-commttee o ISO tea to meet Calcutta ext moth. Avalable ole: Bra BP (001) Prelmary aalyss of the advatages ad lmtatos of a statoary magg Fourer trasform spectrometer, 001 NASA-ODU Amerca Socety for Egeerg Educato (ASEE) Summer Faculty Fellowshp Program, p 48 Che L, Che Q, Zhag Z, Wag X (009) A ovel colormetrc determato of free amo acds cotet tea fusos wth,4- dtrofluorobezee. Joural of Food Composto ad Aalyss, Che Q, Zhao, J, Zhag H, Wag X (006) Feasblty study o qualtatve ad quattatve aalyss tea by ear frared spectroscopy wth multvarate calbrato. Aalytca et Chmca Acta 57, Chug S, Yag, Wag Z (001) Tea ad cacer. Avalable ole: scece/chemstry.html Dez EI, Saz GJM, Pzarro C (004) A evaluato of orthogoal sgal correcto methods for the characterzato of Arabca ad robusta coffee varetes by NIRS. Aalytca et Chmca Acta 514, Ferádez-Ibañez V, Soldado A, Martíez-Ferádez A, de la Roza- Delgado B (009) Applcato of ear frared spectroscopy for rapd detecto of aflatox B 1 maze ad barley as aalytcal qualty assessmet. Food Chemstry 113 (), Fujk H, Sugauma M, Okabe S, Sueoka E, Sueoka N, Fujmoto N, Goto Y, Matsuyama S, Ima K, Nakach K (001) New TNF- releasg hbtors as cacer prevetve agets from tradtoal herbal medce ad combato cacer preveto study wth EGCG ad suldac or tamoxfe. Mutato Research 480/481, Hall MN, Robertso A, Scotter CNG (1998) Near-frared reflectace predcto of qualty, theaflav cotet ad mosture cotet of black tea. Food Chemstry 7, Herrador MA, Gozalez AG (001) Patter recogto procedures for dfferetato of gree, black ad Oolog teas accordg to ther metal cotet from ductvely coupled plasma atomc emsso spectrometry. Talata 53, Hore H, Kohata K (000) Revew-Aalyss of tea compoets by hgh-performace lqud chromatography ad hgh-performace capllary electrophoress. Joural of Chromatography A 881, Hore H, Muka T, Kohata K (1997) Smultaeous determato of qualtatvely mportat compoets gree tea fusos usg capllary electrophoress. Joural of Chromatography A 758, Ioue M, Tajma K, Mzuta M, Iwata H, Iwase T, Mura S, Hrose K, 79

5 Total polypheols by FT-NIR spectroscopy. Sja ad Mshra Hamajma N, Tomaga S (001) Regular cosumpto of gree tea ad the rsk of breast cacer recurrece: follow-up study from the Hosptal-based Epdemologc Research Program at Ach Cacer Ceter (HERPACC), Japa. Cacer Letters 167, ISO (Iteratoal Stadard Orgazato) (1994) Determato of Idvdual Catechs ad Total Polypheols Tea, ISO TC 34/SC 8 N 444 Ja L, Xe LP, Lee AH, Bs CW (004) Protectve effect of gree tea agast prostate cacer: a case-cotrol study southeast Cha. Iteratoal Joural of Cacer 108, L X-L, He Y (008) Evaluato of least squares support vector mache regresso ad other multvarate calbratos determato of teral attrbutes of tea beverages. Food ad Boprocess Techology, press Lupaert J, Zhag MH, Massart DL (003) Feasblty study for the usg ear frared spectroscopy the qualtatve ad quattatve of gree tea, Camella sess (L.). Aalytca et Chemca Acta 487, McGloe VA, Jorda RB, Seelye R, Martse PJ (00) Comparg desty ad NIR methods for measuremet of Kwfrut dry matter ad soluble solds cotet. Postharvest Bology ad Techology 6, McShae MJ, Cote GL (1998) Near-frared spectroscopy for determato of glucose lactate, ad ammoa cell culture meda. Appled Spectroscopy 5, Nakach K, Matsuyama S, Myake S, Sugauma M, Ima K (000) Prevetve effects of drkg gree tear o cacer ad cardovascular dsease: epdemologcal evdece for multple targetg preveto. Bofactors 13, Neuwoudt HH, Pretorus IS, Bauer FF, Nel DG, Pror BA (006) Rapd screeg of the fermetato profles of we yeasts by Fourer trasform frared spectroscopy. Joural of Mcrobologcal Methods 67, Paradkar MM, Irudayaraj J (00) A rapd FTIR spectroscopc method for estmato of caffee soft drks ad total methylxathes tea ad coffee. Joural of Food Scece 67 (7), Pk J, Naczk M, Pk D (1998) Evaluato of the qualty of froze mced red hake: use of Fourer trasform frared spectroscopy. Joural of Agrcultural ad Food Chemstry 46, Sadasvam S, Mackam A (1996) Bochemcal Methods, New Age Iteratoal Pty., Ltd. Publshers, New Delh, pp Satos Jr. VO, Olvera FCC, Lma DG, Petry AC, Garca E, Suarez PAZ, Rubm JC (005) A comparatve study of desel aalyss by FTIR, FTNIR ad FT-Rama spectroscopy usg PLS ad artfcal eural etwork aalyss. Aalytca et Chmca Acta 547, Sawalha SMS, Arráez-Romá D, Segura-Carretero A, Ferádez-Gutérrez A (009) Quatfcato of ma pheolc compouds sweet ad btter orage peel usg CE MS/MS. Food Chemstry 116, Schulz H, Egelhardt UH, Weget A (1999) Applcato of NIRS to the smultaeous predcto alkalods ad pheolc substace gree tea leaves. Joural of Agrcultural ad Food Chemstry 475, Setawa VW, Zhag ZF, Yu GP, Lu QY, L YL, Lu ML, Wag MR, Guo CH, Yu SZ, Kurtz RC, Hseh CC (001) Protectve effect of gree tea o the rsks of chroc gastrts ad stomach cacer. Iteratoal Joural of Cacer 9, Sja VR (009) Gree tea powder ad graules: process techology, storage ad qualty evaluato. Upublshed PhD thess, Ida Isttute of Techology, Kharagpur, Ida, 00 pp Sja VR, Mshra HN (009) FT-NIR spectroscopy for caffee estmato stat gree tea powder ad graules. LWT - Food Scece ad Techology 4, Shbata K, Moryama M, Fukushma T, Kaetsu A, Myazak M, Ue H (000) Gree tea cosumpto ad chroc atrophc gastrts: a cross-sectoal study a gree tea producto vllage. Joural of Epdemology 10, Togar N, Kobayash A, Ashma T (1995) Patter recogto appled to gas chromatographc profles of volatle compoet three tea categores. Food Research Iteratoal 8, Wllams P, Norrs K (001) Near-Ifrared Techology the Agrcultural ad Food Idustres, Amerca Assocato of Cereal Chemsts, St. Paul, MN, USA, 167 pp Woodcock T, Dowey G, O Doell CP (008) Better qualty food ad beverages: the role of ear frared spectroscopy. Joural of Near Ifrared Spectroscopy 16 (1), 1-9 Yag CS, Malakal P, Meg X (00) Ihbto of carcogeess by tea. Aual Revews Pharmacology ad Toxcology 4, 5-54 Zhag D, Kuhr S, Egelhardt UH (199) Ifluece of catechs ad theaflavs o astrget taste of black tea brews, Zetschrft für Lebesmttel- Utersuchug ud -Forschug A (Europea Food Research ad Techology) 195, Zhag MH, Lupaert J, Xu QS, Massart DL (004) Determato of total atoxdat capacty gree tea by NIRS ad multvarate calbrato. Talata 6, 5-35 Zuo YG, Che H, Deg YW (00) Smultaeous determato of catechs, caffee ad gallc acds gree, Oolog, black ad pu-erh teas usg HPLC wth a photodode array detector. Talata 57,