当前位置:首页 期刊杂志

基于电子鼻的花生有害霉菌种类识别及侵染程度定量检测

时间:2024-05-24

沈 飞,刘 鹏,蒋雪松,邵小龙,万忠民,宋 伟



基于电子鼻的花生有害霉菌种类识别及侵染程度定量检测

沈 飞1,2,刘 鹏3,蒋雪松3,邵小龙1,2,万忠民1,2,宋 伟1,2

(1. 南京财经大学食品科学与工程学院,南京 210023; 2. 江苏省现代粮食流通与安全协同创新中心,南京 210023; 3. 南京林业大学机械电子工程学院,南京 210037)

针对花生霉变传统分析方法操作繁琐、时效性差等不足,该研究拟利用电子鼻气体传感技术建立起花生有害霉菌污染的快速检测方法。辐射灭菌花生籽粒分别接种5种谷物中常见有害霉菌(黄曲霉3.17、黄曲霉3.395 0、寄生曲霉3.395、寄生曲霉3.012 4和赭曲霉3.648 6),并于26 ℃、80%相对湿度条件下储藏9 d至严重霉变。利用电子鼻气体传感器获取不同储藏时期(0、3、6、9 d)花生样品的整体挥发性气味信息。最后,结合多元统计分析方法对电子鼻传感器响应信号进行特征提取,建立了花生中有害霉菌污染程度的定性定量分析模型。结果显示,主成分分析法(principal component analysis,PCA)可成功区分不同霉菌侵染程度的花生样品,线性判别分析(linear discriminant analysis,LDA)模型对样品不同储藏天数判别的准确率均达到或接近100%。花生中菌落总数的偏最小二乘回归分析(partial least squares regression,PLSR)模型的预测决定系数和预测相对均方根误差分别达到0.814 5和0.244 0 lg(CFU/g)。结果表明,应用电子鼻技术快速检测储藏期间花生霉变状况具有一定可行性,可为利用气味信息实现粮食霉菌污染的在线监测提供理论参考。

农作物;模型;特征提取;电子鼻;花生;有害霉菌;快速检测

0 引 言

花生富含蛋白质、油脂以及人体必需氨基酸等营养物质,深受消费者喜爱。然而花生极易侵染黄曲霉、寄生曲霉、赭曲霉等有害霉菌而发霉[1-2],产生的黄曲霉毒素(aflatoxin,AFT)、赭曲霉毒素(ochratoxin,OT)等致癌物质,严重威胁人畜健康。目前,花生霉变的检测方法主要有平板计数法、薄层色谱法(thin layer chromatography,TLC)[3]、高效液相色谱法(high performance liquid chromatography,HPLC)[4]和酶联免疫吸附法(enzyme linked immunosorbent assay,ELISA)[5]等。虽然这些方法检测精度较高,但存在操作繁琐、时效性差、成本高,难以满足现场快速检测的需求。因此,寻找一种快速、准确的霉变分析方法,对维护食品安全和消费者身心健康具有重要意义。

目前霉变的相关快速检测技术已有部分报道,如近红外[6]、中红外[7-8]、紫外荧光[9]、机器视觉[10]和电子鼻等。其中,电子鼻作为一种快速、无损的气味检测手段,无需对样品进行复杂的前处理,已广泛用于水果[11]、饮料[12]、酒类[13]、肉类[14]等各类食品的质量评估中,在粮食品质[15]、虫害[16]、新鲜度[17]和霉变程度[18]等方面也有了一些成功的应用。在花生品质分析方面,惠国华等[19]采用电子鼻和随机共振数据分析方法对自然储藏条件下红皮花生的霉变程度进行了快速预测。史文青等[20]应用PEN3型电子鼻对新鲜与烘烤花生的挥发性物质进行了比较研究,确定烘烤后花生气味成分变化主要体现在吡嗪类化合物。Wei等[21]研究表明带壳与去壳花生储藏期间的总酸、过氧化值含量与电子鼻响应信号存在高度相关性,偏最小二乘回归分析(partial least squares regression,PLSR)模型的决定系数2均大于0.80。Jensen等[22]采用电子鼻结合PLSR法对贮藏期花生中的自由基与己醛含量进行了预测,结果显示花生品质与传感器响应信号存在较高相关性。诸多研究显示,应用电子鼻技术分析花生理化指标的研究较多,针对花生霉变的报道不多,且主要集中在初步定性判别阶段,未对花生受不同霉菌感染从而产生的差异开展深入研究,难以准确及有效反映花生霉菌污染的状况。

因此,本研究拟以接种不同种类有害霉菌的花生籽粒为研究对象,应用电子鼻气体传感器阵列获取不同储藏阶段(0、3、6、9 d)花生样品的挥发性气味信息,结合多元统计分析方法建立花生侵染单一霉菌及多种霉菌侵染程度的快速分析模型,并通过气味信息预测花生中菌落总数含量,同时实现花生受霉菌污染程度的定性与定量分析,并对浸染不同霉菌的花生样品的差异进行了研究,进一步验证电子鼻分析技术用于粮食霉变预警的可行性,为开发粮食检测专用设备提供参考。

1 材料与方法

1.1 试验材料

1.1.1 试验样品

花生,购于当地超市,筛选外观良好、形态大小一致、无异味的样品,经Co-60辐射(剂量:15 kGy)灭菌后装入无菌塑料密封袋,置于4 ℃环境下,备用。

1.1.2 霉菌孢子悬浮液制备

5种花生中常见有害霉菌:黄曲霉(A)3.17、黄曲霉(A)3.395 0、寄生曲霉(A)3.395、寄生曲霉(A)3.012 4、赭曲霉(A)3.648 6,均购于中国北京北纳创联研究院。分别将霉菌置于马铃薯葡萄糖琼脂(potato dextrose agar,PDA)培养基上,于26 ℃、80%相对湿度(relative humidity,RH)条件下培养7 d。采用无菌水冲洗培养基表面菌丝,收集孢子悬浮液于50 mL锥形瓶中,参照GB/T4789.2-2010法[23]统计菌落总数,并通过无菌水调整孢子浓度至1×105CFU/mL,4 ℃冷藏,备用。

1.2 试验方法

称取120份灭菌花生样品(50 g/份),置于直径为120 mm的培养皿中,通过移液器分别将每粒花生表面接种10L的黄曲霉3.17孢子悬浮液,然后于26 ℃、80% RH培养箱中储藏9 d,在第0、3、6、9天各取6份样品进行分析,并分别对其余4种孢子悬浮液进行相同处理(总计4×6×5=120份样品)。

采用Fox 3000型电子鼻(法国Alpha Mos公司)检测样品挥发性气味信息,该仪器主要包括顶空全自动进样器、12根金属氧化物传感器与AlphaSoft软件操作系统3部分。检测原理:由于气体与传感器接触发生氧化还原反应,改变传感器导电材料的导电性,并以电阻变化值输出信号,即通过相对电导率(/0)反映传感器响应信号的变化,其中和0分别为传感器吸附样品气、零级气体(经活性炭和硅胶过滤后的空气)后的电导率值[24]。为使花生中霉菌及其代谢产物分布更加均匀,采用粉碎机将样品粉碎。检测前需将样品置于室温下(23±1)℃,2 h后分别称取每份样品2.5 g,置于20 mL顶空瓶中,进行电子鼻检测。样品重复测定3次,取平均值进行分析。花生中菌落总数的测定方法同上。

1.3 数据分析

先运用主成分分析(principal component analysis,PCA)提取电子鼻传感器信号响应值的主成分得分,分析样品变化趋势;再通过线性判别分析(linear discriminant analysis,LDA)对4个不同储藏阶段的样品进行区分;最后通过偏最小二乘回归分析(partial least squares regression,PLSR)对样品中菌落总数进行预测分析。评估PLSR建模性能指标有:模型决定系数(correlation coefficient of determination,2)、建模均方根误差(root mean squared error of calibration,RMSEC)、预测均方根误差(root mean squared error of prediction,RMSEP)、交互验证均方根误差(root mean squared error of cross validation,RMSECV)和相对分析偏差(residual predictive deviation,RPD),其中RPD为预测集标准偏差与RMSEP的比值。以上分析均在Matlab 2014a中进行。

2 结果与分析

2.1 电子鼻气体传感器响应信号分析

图1为受霉菌侵染花生样品的12个不同型号电子鼻气体传感器响应信号随时间变化的响应曲线图。由图1可知,各传感器初始信号平稳,相对电导率(/0)值均为1,随着传感器表面吸附物质增加,每个传感器的响应值均出现不同程度的变化,在13 s附近时达到峰值,随后各传感器响应值逐渐趋于稳定。除LY2/LG、LY2/gCT与LY2/AA外,其余9个传感器的/0值变化较为明显,其中T30/1和LY2/G最为突出。参考各传感器的敏感物质类型可知[25],T30/1对有机化合物较为敏感,LY2/G对胺类化合物与碳氢化合物较为敏感,表明霉变花生中此类物质含量可能较高。结果初步显示,电子鼻各传感器对侵染霉菌花生样品的挥发性物质有明显响应,且不同传感器的响应信号峰值差异明显。

2.2 花生样品菌落总数与霉变程度划分

依据相关研究,根据样品中菌落总数高低将花生样品分为健康(<2.7 lg(CFU/g))、霉变([2.7~4] lg(CFU/g))和重度霉变(>4 lg(CFU/g))3类[26]。由图2可知,随着储藏时间的延长,各组花生中菌落总数不断增加,样品的霉变程度逐渐变大。图2中5组样品初期的菌落总数略有差异,但均为健康状态,可作为对照组。第3天时,仅赭曲霉3.648 6组达到霉变状态。第6天时,除黄曲霉3.395 0组外,其余4组均达到霉变状态。第9天时,5组样品均达到霉变状态,其中黄曲霉3.17与赭曲霉3.648 6组菌落数增长速度最快(>4 lg(CFU/g)),达到重度霉变程度。尽管不同霉菌的繁殖速率存在差异,但菌落总数整体呈上升趋势,导致样品中霉菌整体的新陈代谢活动越旺盛,致使花生中挥发性物质更加复杂,传感器响应信号也随之产生相应变化,为基于气味信息进行霉变样品的快速鉴别提供了可能。

2.3 PCA及载荷结果分析

PCA通过降维方式将原始变量的主要特性指标提取出来,并保留原始数据的主要信息。图3为侵染霉菌样品不同储藏时期的主成分得分及载荷图。图3中主成分1(PC1)和主成分2(PC2)的累积方差贡献率为99.63%,可反映花生电子鼻挥发物图谱的绝大部分信息,可以此为基础进行后续分析[27]。由图3可知,PCA能较好的区分不同储藏时期的花生样品。0、3和6 d样品的PCA得分呈线性变化,即随着储藏期的延长,霉变样品逐渐沿轴负方向移动,其中第6天的样品能完全区分于0、3 d的样品,说明花生霉变后挥发性物质的种类或含量存在显著变化。另外,第9天的重度霉变样品呈现反向移动,可能与霉菌生长后期代谢产生的大量挥发性次级代谢产物有关。上述变化不受侵染霉菌种类影响,即侵染5种霉菌的样品呈现相同的规律,表明该霉变变化状况信息在花生中具有一定普遍性。当对单一霉菌在储藏9 d进行PCA时也存在类似变化规律。分析表明,花生不同储藏时期产生的挥发性气味存在差异,应用PCA对花生霉变状态进行区分具有可行性。由载荷分析可知,各个传感器在PCA中的贡献大小存在明显差异(图3)。其中,T70/2、LY2/LG、P10/1、T30/1等传感器的相对权重值较大,显示被霉菌侵染的花生籽粒的挥发性组分变化可能主要与氮氧化合物、芳香族化合物和碳氢化合物等有关,同时说明此类传感器在有效鉴别霉变花生样品中具有重要贡献,也为后续专用型电子鼻传感器的开发提供了参考信息。

2.4 LDA结果分析

LDA是模式识别中一种特征提取与降维分析方法。本研究采用Fisher判别法,即利用投影技术将原始数据投影到最佳方向,以实现建模集与验证集的有效区分,判别结果如表1所示。依据储藏时间及受霉菌感染程度的不同,将样品分为4类(0,3,6,9 d)。由表1可知,LDA模型能较好的区分不同储藏时期侵染单一霉菌样品,建模集准确率均为100%,验证集中仅赭曲霉3.648 6组中1个样品被误判,剩余4组样品均能被成功区分。当对侵染5种霉菌的样品进行综合建模时,准确率均达到或接近100%。结果表明,感染不同程度的花生样品可被完全区分。花生在储藏霉变直到后期过程中,由于脂肪、蛋白质等大量霉菌被快速分解,其代谢活动产生大量种类繁多的次级代谢产物(毒素、挥发性物质等),导致样品挥发性气味特征与储藏前期存在一定系统差异,为电子鼻气体传感器提供了判别基础。

表1 侵染霉菌花生的LDA建模分析及验证结果

2.5 PLSR分析

以传感器响应值为自变量,花生中菌落总数(lg (CFU/g))为因变量,选取2/3样品作为建模集,1/3样品作为预测集,采用PLSR法对花生中总菌落数进行预测。为消除噪声、信号漂移等的影响,采用标准正态变换(standard normal variate,SNV)和Savitsky-Golay(15点,2次多项式平滑过滤)法对传感器响应信号进行预处理。

由表2知,PLSR法可较好的预测花生中的菌落总数。建模集中,侵染单一霉菌组模型的建模决定系数(R2)高于0.90,建模均方根误差(RMSEC)低于0.20 lg(CFU/g),且所有模型的偏差(Bias)均小于0.000 4。其中寄生曲霉3.395组的模型结果最优,R2值为0.971 2,RMSEC值为0.072 2 lg(CFU/g)。然而将全部样品进行综合建模时,精度略有降低,R2值为0.820 6。验证集中,对侵染单一霉菌组进行留一交互验证时,交互验证均方根误差(RMSECV)均低于0.30,同样寄生曲霉3.395组的误差最小,而黄曲霉3.17侵染组的RMSECV值最大,为0.248 6 lg(CFU/g)。对单一霉菌进行预测时,除黄曲霉3.395 0侵染组外,其余4组模型的预测决定系数R2和RPD值分别大于0.90、3.0,说明该模型具有一定定量分析的潜力[28]。此外,仅有黄曲霉3.17侵染组的RMSEP值偏大,为0.206 2 lg(CFU/g),其余4组均低于0.20。由于各传感器对气味的灵敏性存在差异(图1),且侵染5种霉菌花生的挥发性成分不一致,导致电子鼻对花生中不同霉菌菌落数的预测误差不同。综上对比,寄生曲霉3.395侵染组预测模型相比剩余4组结果最优,R2、RMSECV与RPD值分别为0.943 6、0.100 2 lg(CFU/g)和4.09。由图4可知,5种霉菌侵染组综合模型的预测精度稍低,R2和RMSEP值分别为0.814 5、0.244 0 lg(CFU/g),明显低于单一霉菌侵染组模型。结果显示所有模型的RPD值均大于2.0,说明这些模型可用于定性分析目的,通过电子鼻预测花生中菌落总数,判断花生是否霉变具有可行性。然而受限于电子鼻系统传感器性能及数量的影响,导致其对侵染不同霉菌样品菌落数的整体预测精度仍有待提升。分析可知,表2显示各模型中参与分析的潜在变量(latent variables,LVs)均≤5,后续分析可通过优化潜在变量数来提升模型精度,此外,进一步研究应当通过优化样品预处理步骤、扩大样品数量及采用自然霉变的样品等方式,达到验证和改善模型性能的目的[29-30]。

表2 花生中霉菌总数PLSR模型预测分析结果

注:R2为建模决定系数;R2为预测决定系数;RMSEC为建模均方根误差;RMSEP为预测均方根误差;RMSECV为交互验证均方根误差;RPD为相对分析偏差。

Note:R2is represents correlation coefficient of determination in calibration;R2is represents correlation coefficient of determination in prediction; RMSEC is represents root mean squared error of calibration; RMSEP is represents root mean squared error of prediction; RMSECV is represents root mean squared error of cross validation; RPD is represents residual predictive deviation.

3 结 论

本文采用电子鼻气体传感器阵列对不同储藏时期侵染霉菌的花生的气味信息进行了检测,并结合多元统计分析方法建立了不同霉变程度样品的定性定量分析模型。PCA及载荷分析结果显示侵染5种霉菌花生样品的传感器响应信号在储藏期间存在一定变化规律,且不同储藏时期的样品均能得到有效区分,霉菌侵染花生引起的挥发性成分变化可能主要在于芳香族化合物、氮氧化合物和碳氢化合物等物质;运用线性判别分析分别对侵染单一霉菌的花生样品及全部样品进行建模,准确率均达到或接近100%;偏最小二回归模型对花生籽粒中菌落总数预测精度较高,5种感染霉菌花生样品综合模型的预测决定系数为0.814 5,预测均方根误差为0.244 0 lg(CFU/g)。上述试验表明,电子鼻技术作为一种快速、高效的气味信息检测手段,用于花生储藏期间霉变状态的鉴别具有一定可行性,可为快速评估花生质量安全提供参考。

[1] 李建辉. 花生中黄曲霉毒素的影响因子及脱毒技术研究[D]. 北京:中国农业科学院,2009.

Li Jianhui. Impact Factor and Detoxification of Aflatoxin in Peanuts[D]. Beijing: Chinese Academy of Agricultural Sciences, 2009. (in Chinese with English abstract)

[2] Afzali D, Fathirad F, Ghaseminezhad S. Determination of trace amounts ofA in different food samples based on gold nanoparticles modified carbon paste electrode[J]. Journal of Food Science and Tenhnology, 2016, 53(1): 909-914.

[3] Kamika I, Takoya L L. Natural occurrence ofB1in peanut collected from kinshasa, democratic republic of congo[J]. Food Control, 2011, 22(11): 1760-1764.

[4] Iqbal S Z, Asi M R, Zuber M, et al. Aflatoxins contamination in peanut and peanut products commercially available in retail markets of Punjab, Pakistan[J]. Food Control, 2013, 32(1): 83-86.

[5] Shukla S. Estimation of aflatoxins in peanut or maize by enzyme linked immunosorbent assay[J]. Bangladesh Journal of Pharmacology, 2016, 11(3): 628-631.

[6] Jiang Jinbao, Qiao Xiaojun, He Ruyan. Use of near-infrared hyperspectral images to identify moldy peanuts[J]. Journal of Food Engineering, 2016, 169: 284-290.

[7] Celiker H K, Mallikarjunan P K, Kaaya A. Characterization of invasion of genuson peanut seeds using FTIR-PAS[J]. Food Analytical Methods, 2016, 9(1): 105-113.

[8] Celiker H K, Mallikarjunan P K, Schmale D, et al. Discrimination of moldy peanuts with reference to aflatoxin using FTIR-ATR system[J]. Food Control, 2014, 44: 64-71.

[9] 王叶群,杨增玲,张绍英,等. 用于污染黄曲霉毒素花生分选的荧光信号研究[J]. 农业工程学报,2016,32(1):187-192.

Wang Yequn, Yang Zengling, Zhang Shaoying, et al. Fluorescent signal characteristics for sorting of peanut contaminated by[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2016, 32(1): 187-192. (in Chinese with English abstract).

[10] Wang J, Yang W W, Walker L T, et al. Identification of peanut pods with three or more kernels by machine vision and neural network[J]. International Journal of Food Engineering, 2014, 10(1): 97-102.

[11] Ciptohadijoyo S, Litananda W S, Rivai M, et al. Electronic nose based on partition column integrated with gas sensor for fruit identification and classification[J]. Computers and Electronics in Agriculture, 2016, 121: 429-435.

[12] Jin Jiaojiao, Deng Shanggui, Ying Xiaoguo, et al. Study of herbal tea beverage discrimination method using electronic nose[J]. Journal of Food Measurement and Characterization, 2015, 9(1): 52-60.

[13] Liu Ming, Han Xiaomin, Tu Kang, et al. Application of electronic nose in Chinese spirits quality control and flavour assessment[J]. Food Control, 2012, 26(2): 564-570.

[14] Lippolis V, Ferrara M, Cercellieri S, et al. Rapid prediction ofA-producing strains ofon dry-cured meat by MOS-based electronic nose[J]. International Journal of Food Microbiology, 2016, 218: 71-77.

[15] Han H J, Dong H, Noh B S. Discrimination of rice volatile compounds under different milling degrees and storage time using an electronic nose[J]. Korean Journal of Food Science and Technology, 2016, 48(2): 187-191.

[16] Lampson B D, Han Y J, Khalilian A, et al. Development of a portable electronic nose for detection of pests and plant damage[J]. Computers and Electronics in Agriculture, 2014, 108: 87-94.

[17] Jiang Jinghao, Li Jian, Zheng Feixiang, et al. Rapid freshness analysis of mantis shrimps () by using electronic nose[J]. Journal of Food Measurement and Characterization, 2016, 10(1): 48-55.

[18] 殷勇,郝银凤,于慧春. 基于多特征融合的电子鼻鉴别玉米霉变程度[J]. 农业工程学报,2016,32(12):254-260.

Yin Yong, Hao Yinfeng, Yu Huichun. Identification method for different moldy degrees of maize using electronic nose coupled with multi-features fusion[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2016, 32(12): 254-260. (in Chinese with English abstract).

[19] 惠国华,倪彧. 基于电子鼻系统的粮食霉变检测方法研究[J]. 中国食品学报,2011,11(5):162-168.

Hui Guohua, Ni Yu. Investigation of moldy grain detection method using electronic nose system[J]. Journal of Chinese Institute of Food Science and Technology, 2011, 11(5): 162-168. (in Chinese with English abstract)

[20] 史文青,薛雅琳,何东平. 花生挥发性香味识别的研究[J]. 中国粮油学报,2012,27(7):58-62.

Shi Wenqing, Xue Yalin, He Dongping. Study on identification of volatile flavor in peanut[J]. Journal of the Chinese Cereals and Oils Association, 2012, 27(7): 58-62. (in Chinese with English abstract).

[21] Wei Zhenbo, Wang Jun, Zhang Weilin. Detecting internal quality of peanuts during storage using electronic nose responses combined with physicochemical methods[J]. Food Chemistry, 2015, 177(15): 89-96.

[22] Jensen P N, Bertelsen G, Van Den Berg F. Monitoring oxidative quality of pork scratchings, peanuts, oatmeal and muesli by sensor array[J]. Journal of the Science of Food and Agriculture, 2005, 85(2): 206-212.

[23] Amalia B. Metal oxide sensors for electronic noses and their application to food analysis[J]. Sensors, 2010, 10(4): 3882-3910.

[24] GB/T4789.2 2010. 食品微生物学检验菌落总数测定[S]. 中国标准出版社,2010.

[25] Xu Guojie, Liu Chunsheng, Liao Caili, et al. Rapid and accurate identification of adulterants via an electronic nose and DNA identification platform: identification of fake velvet antlers as an example[J]. Journal of Sensors, 2016, 2016(23): 1-7.

[26] Celikera H K, Mallikarjunan P K, Kaayab A. Mid-infrared spectroscopy for discrimination and classification of Aspergillus spp. contamination in peanuts[J]. Food Control, 2015, 52: 103-111.

[27] Philip R C N, Paul A T, John F M. Missing data methods in PCA and PLS: score calculations with incomplete observations[J]. Chemometrics and Intelligent Laboratory Systems, 1996, 35(1): 45-65.

[28] Herrmann S, Mayer J, Michel K, et al. Predictive capacity of visible-near infrared spectroscopy for quality parameter assessment of compost[J]. Journal of Near Infrared Spectroscopy, 2009, 17(5): 289-301.

[29] Saevels S, Lammertyn J, Berna A Z, et al. Electronic nose as a non-destructive tool to evaluate the optimal harvest date of apples[J]. Postharvest Biology and Technology, 2003, 30(1): 3-14.

[30] Shen Fei, Ying Yibin, Li Bobin, et al. Prediction of sugars and acids in Chinese rice wine by mid-infrared spectroscopy[J]. Food Research International, 2011, 44(5): 1521-1527.

Recognition of harmful fungal species and quantitative detection of fungal contamination in peanuts based on electronic nose technology

Shen Fei1,2, Liu Peng3, Jiang Xuesong3, Shao Xiaolong1,2, Wan Zhongmin1,2, Song Wei1,2

(1.,,210023,; 2.,210023,; 3.,,210037,)

Current methods for fungi contamination determination in peanuts are usually labor-intensive and time-consuming. In this paper, a new method for rapid detection of the contamination by harmful fungi species in peanut kernels based on electronic nose (E-nose) technology was investigated. Peanut samples were firstly irradiated by Co-60 gamma radiation with a dose of 15 kGy to kill all fungi on or within kernels. After irradiation, clean and sterile peanuts were placed in moist chambers and inoculated with 5 different spore suspensions of, which were A.3.17, A.3.395 0, A.3.395, A.3.012 4 and A.3.648 6, the former 3 of which were aflatoxin (AFT) producer, and the latter one was ochratoxin (OT) producer. Spore suspensions were prepared by blending the 7-day old colonies cultured on potato dextrose agar (PDA) with ultrapure sterilized water. Initial spore concentration was about 5 log (CFU/mL), and then 10L spore suspension was dropped onto individual peanut sample by a pipette. All infected samples were stored at 26 ℃ and 80% relative humidity (RH) for 9 d until all peanut samples were covered with a mass of fungi. Subsequently, the E-nose (Fox 3000, Alpha Mos) was used for the collection of volatile odor information from peanut samples stored for 0, 3, 6 and 9 d, respectively. Finally, response signals of 12 E-nose sensors were extracted by multivariate statistical analysis method. Qualitative and quantitative models for the determination of harmful fungi contamination in peanuts were established. The principal component analysis (PCA) results showed that peanut samples with different storage days could be successfully discriminated for different fungal infection levels. Loading analysis of E-nose sensors indicated that the sensors of T70/2, LY2/LG, P10/1, T30/1 were found to be more sensitive than other sensors. These sensors might play an important role in the discrimination of samples, which provided a reference for the development of special-purpose sensor systems for peanut samples in future. The changes in volatile compounds of infected peanut samples could be mainly attributed to oxynitride, hydrocarbon and aromatic compounds. For the classification of peanut samples with different infection levels, the correct rate of 100%(or approaching) was obtained by linear discriminant analysis (LDA) models. The results also verified the possibility of discriminating peanuts infection by different fungi species. In addition, good correlation between E-nose signals and colony forming units in peanut samples was obtained by partial least squares regression (PLSR) analysis models. The coefficient of determination for the prediction set (R2) and the root mean square error of prediction (RMSEP) for the prediction models were 0.814 5 and 0.244 0 lg (CFU/g), respectively. Both LDA and PLSR methods were proven to be effective in the discrimination/ quantification of fungi contamination in peanuts. The results indicate that E-nose technology can be used as a feasible and reliable method for the determination of peanut quality during the storage, which can provide the theoretical reference for rapid detection of mold contamination during grain storage using volatile odor information.

crops; models; feature extraction; electronic nose; peanuts; harmful fungi; rapid detection

10.11975/j.issn.1002-6819.2016.24.040

TS255.7; S379

A

1002-6819(2016)-24-0297-06

2016-07-01

2016-11-20

国家自然科学基金青年基金(31301482);江苏省青年自然科学基金(BK20131007);粮食公益性行业科研专项(201513002-5);江苏高校优势学科建设工程资助项目(PAPD)(2014-124)

沈 飞,男,博士,副教授,硕士生导师,主要研究方向为粮食储藏与品质无损检测。南京 南京财经大学食品科学与工程学院,210023。Email:shenfei0808@163.com

免责声明

我们致力于保护作者版权,注重分享,被刊用文章因无法核实真实出处,未能及时与作者取得联系,或有版权异议的,请联系管理员,我们会立即处理! 部分文章是来自各大过期杂志,内容仅供学习参考,不准确地方联系删除处理!