Preliminary study on the use of near infrared hyperspectral imaging for quantitation and localisation of total glucosinolates in freeze-dried broccoli

TL;DR: In this article, the use of hyperspectral imaging to quantify and localise total glucosinolates in florets of a single broccoli species has been examined Two different spectral regions (vis-NIR and NIR), a number of spectral pre-treatments and different mask development strategies were studied to develop the quantitative models.

About: This article is published in Journal of Food Engineering.The article was published on 2014-04-01 and is currently open access. It has received 28 citations till now. The article focuses on the topics: Hyperspectral imaging.

Summary

1. Introduction

• The aim of this study was to evaluate the potential of hyperspectral imaging technology for the quantitative screening and localisation of total glucosinolates in freeze-dried broccoli.
• Since predictive models developed on freeze-dried powders by conventional NIR spectrometers may not be transferred directly to hyperspectral imaging datasets, a new predictive model must be generated using an actual hyperspectral imaging system on homogeneous, freeze-dried broccoli powders after which it may be applied to hyperspectral images of intact broccoli for localisation and quantitation of total glucosinolates.
• To their knowledge, this is the first time that this analytical tool has been applied to broccoli for these purposes.

2.2. Hyperspectral imaging analysis

• Data were recorded in units of reflectance and saved in ENVI header format using the instrument acquisition software (Spectral Scanner; DV Optics, Padua, Italy).
• When using System 1, only spectral data in the 450 -900 nm regions were used in data analysis due to reduced efficiency of the light source and CCD in wavelength regions outside this range.
• In the case of System 2, the spectral range was attenuated to 950 -1650 nm for similar operational reasons.

2.3. Data processing and analysis

• Data treatment and quantitative model development was carried out using Matlab (R2010b; The Math Works, Inc. USA).
• For each hyperspectral image, regions of interest (ROIs) of approx.
• 3 cm diameter were selected using an interactive selection tool available in the acquisition software ('ROI tool') and 500 pixels were randomly-selected within each ROI.
• Spectral data were pre-treated using the standard normal variate (SNV) transform to diminish the effects of light scatter.
• Finally, quantitative calibrations were developed by partial least squares (PLS) regression using total glucosinolates as the dependent (Y) variable and pixel spectra as the independent (X) variables.

2.4 Prediction map

• Slices of whole freeze-dried broccoli were scanned in the NIR zone (950-1650 nm) to apply the previously constructed model and identify the glucosinolate allocation.
• Prior to the quantitative analysis, a thresholding rule method was applied to the broccoli images to isolate the broccoli from other parts of image.
• An image was generated using the maximum reflectance value of each pixel spectrum in a raw image.
• A threshold of 0.45 reflectance units was set analysing the corresponding histogram and drawing a tentative mask image in an iterative process.
• SNV was applied to minimise the effects of scattering in the mask created and then the PLS model was applied.

3.2. Prediction map

• Their potent odour and pronounced taste suggests a role in herbivore and microbial defence.
• Deposition in external plant parts, confirmed spectroscopically in this work, would be the optimal location for these purposes.

4. Conclusion

• Two different spectral regions (vis-NIR and NIR) were studied to develop the quantitative models.
• Better results were obtained using the 950-1650 nm wavelength range and subsequent analyses were therefore carried out using this spectral zone.
• The procedure demonstrates potential for the quantitative screening and location of total glucosinolates in broccoli using the 950-1650 nm wavelength range.
• Nevertheless, a comprehensive study should be made in order to evaluate all other relevant sources of variability in the complete development of these models.
• Such a study would entail several years work but the results reported herein suggest the viability of obtaining useful results from such an undertaking.

Figures

Fig. 3

Fig. 4

Fig. 5

Fig. 1

Fig. 2

Frequently asked questions

Q1. What are the contributions mentioned in the paper "Preliminary study on the use of near infrared hyperspectral imaging for quantitation and localisation of total glucosinolates in freeze-dried broccoli" ?

1 The use of hyperspectral imaging to ( a ) quantify and ( b ) localise total glucosinolates in 2 florets of a single broccoli species has been examined. Two different spectral regions 3 ( vis-NIR and NIR ), a number of spectral pre-treatments and different mask development 4 strategies were studied to develop the quantitative models. The procedure demonstrates potential for the 7 quantitative screening and localisation of total glucosinolates in broccoli using the 9508 1650 nm wavelength range.