Follow us
Comparing Hyperspectral Imaging and Near IR spectroscopy for detection of Salmonella and other bacteria in Milk, Juice and Poultry

Milk is rich in its ingredients and contains fat, protein minerals, vitamins, carbohydrates and tens of thousands of other ingredients. It is a perfect environment for bacterial cultures such as Salmonella to develop due its rich nutritional nature. Unprocessed milk that has not been heat treated or pasteurized can host a variety of bacterial cultures that could cause Gastroenteritis and other intestinal diseases (Reference 1). Figure 1 shows the variety of bacterial cultures that could exist in milk.


Figure 1: Different bacteria present in Milk

Another source of bacterial infection is poultry. The number of cases due to Salmonella sickness in the United States exceeds one million per year and out of this, 200,000 cases are originated from poultry alone (Reference 2). The bacterial infection can also occur in commercial fruit pulp such as pine-apple. In addition, orange juice and mango juice could also get infected by bacteria (Reference 3).

The analytical methods of detection of Salmonella such as Polymerase Chain Reaction (PCR) require sample preparation, colony isolation and confirmation. Considerable amount of liquid and solid media need to be consumed and reagents are required. Not to mention, the turn-around time of the test approximating 5 to 7 days.

A faster, non-destructive, non-invasive method of testing which does not require chemical reactions are required for testing and identification of Salmonella. There are two candidates for this kind of fast testing. These two methods are hyperspectral imaging and near IR Spectroscopy (NIRS). Hyperspectral Microscopic Imaging (HMI) which uses an Acousto-Optic Tunable Filter (AOTF) and a microscope to look at individual bacterial cells from micro-colony on agar plates and the use of Principle Component Analysis (PCA) to detect and identify different types of bacteria (Reference 2). A picture of the setup is shown in figure 2.

Odoo CMS - a big picture

    Figure 2: AOTF microscopic imaging system to detect Salmonella (Reference 2)

The AOTF microscopic system classifies potential pathogens using only a few cells and analyses the data using Chemometrics.
The second option for identifying Salmonella and possibly distinguishing it from other samples infected with the E-Coli is Near-IR Spectrocopy (NIRS). The absorption in the near-IR region between 780-2500 nm consists of overtones and combination bands which are broader and have lower resolution as compared to the Mid-IR region where the fundamental absorptions exist. However there is no need for sample preparation and the measurements will take a few seconds as compared to a few days. Due to the complexity of spectra and in order to distinguish between overlapping absorption bands, NIRS requires chemometrics methods to analyze the data. Several different bonds such as C-H, N-H, O-H and C=O can be detected using NIRS. In one research work, a near-IR spectrometer working in the range 900-1700 nm was used to discriminate between salmonella infected milk samples and non-infected milk samples (Reference 1). After performing a PLS-DA (Partial Least Square-Discriminant Analysis) examination of the second derivative of spectra, the following plot was obtained as shown in Figure 3.

Continue Reading...
Odoo CMS - a big picture

Figure 3: PLS -DA method  distinguishes between contaminated and uncontaminated samples in Milk after NIRS measurement.

In the plot above, class 1 and class 2 refer to calibration samples (18 contaminated and 18 clean). Class 1 val  and 2 val  refer to validation samples (8 contaminated, 8 clean). It is clear that the two classes have been separated very well using this method and  proves that the NIRS makes real-time monitoring possible and can reduce the processing time significantly.

In another study, the goal was to distinguish between the fruit juice samples infected with Salmonella and the ones infected with E-Coli (Reference 3). Partial Least Square (PCA) method   could not distinguish between the two types of bacteria but when PLS-DA was used the two types were clearly separated. The measurement used an FT-NIR spectrometer in the range 1000-2500 nm to obtain the spectra. FT-NIR uses a Fourier transform technique and a Near IR Halogen Tungsten lamp to do the measurements. Figure 4 shows the comparison between the PCA method and the PLS-DA (Partial Least Square-Discriminant Analysis) method. It seems that the former has failed and the latter has succeeded to distinguish between Salmonella and E-Coli.

Odoo CMS - a big picture

PLS-D

Odoo CMS - a big picture

Figure 4: PCA and PLS-DA comparison

So it becomes apparent that a proper chemometrics technique is required for classification and not every method will be able to make a distinction between samples of different nature.

Comparison of different methods

In this blog, three different techniques were reviewed. Hyperspectral Microscopic Imaging (HMI), NIRS using a diffraction grating based near-IR spectrometer and finally NIRS using FT-NIR. All the three methods used chemometrics to identify the bacteria. However as far as the cost of the hardware is concerned, both the HMI and FT-NIR are very expensive instruments and their price is in tens of thousands of dollars. The diffraction grating based NIR spectrometers can be as low as $2000. Of course the chemometrics part of the work is altogether separate and either needs to be done on a PC with a statistical software such as MatLab-PLS tool box or an app needs to be developed that uses cloud based statistical tools such as Python based statistical libraries.

Allied Scientific Pro offers the Nirvascan spectrometer that comes in different models (Reflectance model for solids, Transmission model  for liquids and fiber optic model which requires an external source). Although the spectrometer is diffraction grating based, it uses the Digital Light Processing (DLP) technology and a single element InGaAs detector which considerably reduces the price. A pic of each three model is shown in figure 5.

Figure 5: Different models of Nirvascan spectrometer

For more information, refer to the following link.

https://alliedscientificpro. com/nirvascan

References:

1-   Fast Discrimination of Milk Contaminated with Salmonella sp. Via Near-Infrared Spectroscopy, J.M. Pereira et.al, Food Anal. Methods (2018).

2-  Classification of Salmonella Serotypes with Hyperspectral Microscope Imagery, B. Park et.al, Ann Clin Pathol 5(2):1108 (2017)

3-  The use of near infrared spectroscopy and multivariate techniques to differentiate Escherichia coli and Salmonella Enteritidis inoculated into pulp juice, A. Marques et.al, Journal of Microbiological methods 93 (2013) 90-94

Detection of fraudulent fish using NIR spectroscopy and Chemometrics