Method of Optical Diagnostics of Grain Seeds Infected with Fusarium

Featured Application

The results of this work can be used in breeding, seed production, phytopathology and agricultural engineering.

Abstract

Optical sensors have shown good capabilities for detecting and monitoring plant diseases, including fusariosis. The spectral characteristics of the excitation and luminescence of wheat, oat and barley seeds were measured using a diffraction spectrofluorimeter in the range of 180–700 nm. It was found that during infection, the spectral density of the absorption capacity increases and the curve η_e(λ) shifts upwards in the range of 380–450 nm. The shift to the left is also noticeable for the wheat and barley spectra. The photoluminescence flux at λ_e = 232 nm increased by 1.71 times when oat seeds were infected, by 2.63 times when wheat was infected and by 3.14 times when barley was infected. The dependences of the infection degree on the photoluminescence flux are statistically and reliably approximated by linear regression models with determination coefficients R² = 0.83–0.95. The method of determining the degree of infection can include both absolute measurements of photoluminescence flux in the range of 290–380 nm and measurements of the flux ratios when excited by radiation of 232 nm and 424 nm for wheat and 485 nm for barley. An optoelectronic device for remote monitoring can be designed in order to implement the methodology for determining the degree of infection of agricultural plant seeds.

1. Introduction

Infections of the grain plants Fusarium are serious problems all over the world due to the mycotoxins potentially produced by these fungi. Infections of grain crops, especially fusarium, are a serious, ever-growing problem. Crop loss of up to 30% may indicate the enormous consequences of this disease. This has led to the need to develop tools for the early detection of infections. Common diagnostic methods are visual, as well as mycological analysis, enzyme immunoassay, real-time polymerase chain reaction (PCR) and quantitative digital PCR, determination using markers based on intergenic spacers. Optical sensors have shown good opportunities for improving the detection and monitoring of plant diseases. Such sensors can be built on infrared thermography, chlorophyll fluorescence imaging and hyperspectral imaging (HSI) [1].

The occurrence of fusarium head disease can be detected by spectral analysis (400–1000 nm) before harvesting. Wheat plants were analyzed using a hyperspectral imaging system under laboratory conditions [2,3]. The use of HSI for the detection of fusarium-damaged kernels in Canadian wheat samples was investigated [4]. Transmission spectral images in both the visible and near-infrared ranges from infected nuclei were recorded in single wheat grains. Polder [5] found that spectral images in the near-infrared (NIR) range worked much better than spectral images in the visible range. Classification models based on selected texture parameters have been developed to distinguish between infected and healthy wheat grains [6]. A specific classification index of fusariosis for the detection of this disease in wheat was proposed. Hyperspectral microscopy images of wheat spikelets were used as a data source [7].

The possibility of using hyperspectral imaging to detect the content of deoxynivalenol and damage of single oat kernels by fusarium was presented [8]. The absorption spectra of deoxynivalenol in the near-infrared range of single wheat kernels were studied. The absorption spectra of the NIR were recorded in the range of 350–2500 nm [9]. In the study [10], visible and near-infrared spectroscopy and computer vision methods were combined to simulate the online distinction between normal and deoxynivalenol-contaminated wheat grains.

The difference in spectral reflection between healthy and infected wheat plants which were infected with Puccinia striiformis was studied [11]. The degree of damage from streak rust, brown rust and the spotting of Septoria tritici was evaluated in vivo, as well as for the selection of informative wave ranges based on images obtained using a multispectral camera integrated into a ground platform [12].

The use of hyperspectral NIR visualization and multidimensional image analysis on a potato dextrose field made it possible to visualize radial growth rings on images of the main components. Hyperspectral imaging of NIR in combination with multidimensional image analysis is an effective tool for assessing the growth characteristics of microbes [13].

The hyperspectral image in the spectral range of 900–2500 nm was studied for its ability to identify corn grains inoculated with aflatoxigenic fungus AF13 from healthy nuclei and nuclei inoculated with non-aflatoxigenic fungus AF36 [14], and also for recognizing different varieties of corn seeds [15]. Thermal (8–13 µm) and hyperspectral images in the visible, near-infrared and short-wave infrared ranges were used to develop a method for the early detection of biotic stresses caused by a species of fungi belonging to the genus Alternaria for oilseed rape [16]. A fast, non-destructive and accurate HSI-based method has been developed to monitor the growth of spoilage-causing fungi on stored brown rice. The HSI system was used to obtain images of the reflectivity of samples in the visible and near-infrared wavelength range of 400–1000 nm [17]. A step-by-step HSI method for the detection of bacterial burning of rice shells was proposed. The method can potentially serve as a tool for the high-performance detection of plant stem diseases in field conditions [18]. It has been proven that spectroscopy in the middle and near-infrared range is effective in monitoring the fungal contamination of champignons [19].

The early non-destructive detection of damage during cooling in eggplants was investigated using spectroscopy. The results indicate the promising potential of NIR spectroscopy to provide non-invasive, fast and reliable detection of damage in eggplants [20]. The possibility of using spectroscopy in the mid-infrared or near-infrared range to predict the vitamin C content in plum powder samples has been shown [21]. Hyperspectral imaging (380–1020 nm) and machine learning were used to develop a technique for detecting various stages of the disease (asymptomatic, early, intermediate and late stages) of powdery mildew in zucchini [22].

Computer vision technology was developed in combination with an artificial neural network to identify frogeye soybeans, moldy soybeans, wormy soybeans and damaged soybeans [23].

At the same time, reflective infrared spectroscopy requires expensive, high-precision equipment, so luminescent spectroscopy of the ultraviolet and visible range can be used as an alternative to it. A deeper and longer interaction of radiation and biological tissue occurs during the excitation and emission of luminescence than during reflection. Unfortunately, very little attention is paid to optical photoluminescent diagnostics of plant diseases in the modern scientific literature. For example, the article [24] describes the spectral luminescent diagnostics of citrus fruits.

The purpose of this study is to take up the spectral luminescent properties of healthy and infected seeds and to develop a technique of optical diagnostics for fusariosis of grain plants.

2. Materials and Methods

One of the most common and dangerous diseases for plants—fusariosis—was selected for research, the causative agents of which are fungi of the genus Fusarium. Seeds of winter wheat “Irishka No. 172”, barley “Moskovsky 86” and oats “Zalp” were used as biomaterial.

The degree of seed infection was determined by external signs. The mass of the average sample was 2.0 ± 0.1 kg; the mass of the selected sample was at least 25 g. According to this method, grains infected with fusariosis were selected according to their appearance: the shape and structure of the grain were puny, they had a strongly depressed groove; the characteristic of the grain surface—spots and plaque; the structure of the endosperm—a significant loss of vitreous, the endosperm was loose; the color of the embryo was a dark-colored embryo (brown), and the fungus plaque was present on the germ part and in the groove. Additionally, the method of thin-layer chromatography was used to detect the T-2 toxin by fluorescence in long-wave ultraviolet light (365 nm) after treatment with an alcoholic solution of sulfuric acid, followed by heating at 100–105 °C. This method can detect up to 100 ng of T-2 toxin in a spot of infected grain [25].

The spectral characteristics of seed excitation and luminescence were measured using a previously developed technique [26] based on a Fluorat-02-Panorama spectrofluorimeter (manufactured by Lumex, Russia) with PanoramaPro software. The spectrofluorimeter includes an optical circuit with a radiation source—a pulsed xenon lamp and a radiation receiver—a photoelectronic multiplier. The lamp operates in the mode of short pulses with a repetition frequency of 25 Hz. Monochromators with a concave diffraction grating are used to isolate the necessary spectral range.

In the first stage, the excitation (absorption) spectrum of the seeds of η_e(λ) was measured by synchronous scanning with spectrofluorimeter monochromators. Analysis of the obtained spectra of η_e(λ) made it possible to determine the wavelengths (or spectral regions) with which to excite photoluminescence. In the second stage, the seeds were excited by radiation, and the photoluminescence spectra φ_l(λ) were measured using a spectrofluorometer. According to the measurement results, statistical processing was carried out, where averaging was carried out over 250 spectra. The integral parameters of the H and Φ spectra were calculated in the PanoramaPro program.

$$\mathrm{H}={{\displaystyle \int }}_{{\mathsf{\lambda }}_1}^{{\mathsf{\lambda }}_2}{\mathsf{\eta }}_{\mathrm{e}}\left(\mathsf{\lambda }\right)\mathrm{d}\mathsf{\lambda },$$

(1)

• η_e(λ)—spectral characteristics of excitation,
• λ₁…λ₂—limits of operating spectral range of excitation,

$$\mathsf{\Phi }={{\displaystyle \int }}_{{\mathsf{\lambda }}_1}^{{\mathsf{\lambda }}_2}{\mathsf{\phi }}_l\left(\mathsf{\lambda }\right)\mathrm{d}\mathsf{\lambda },$$

(2)

• φ_l(λ)—spectral characteristics of photoluminescence,
• λ₁…λ₂—limits of operating spectral range of photoluminescence.

This parameter is the photoluminescence flux, expressed in relative units.

Statistical processing of the results was carried out using generally accepted formulas.

3. Results

Two hundred and fifty measurements were carried out with the simultaneous scanning of infected and uninfected seeds. The average results of the measurements of the oat seeds are presented as an example in Figure 1. The values of spectral densities of excitation and luminescence are given in relative units (r. u.) and reduced to 100 units at the maximum sensitivity of the spectrofluorimeter.

It can be seen from Figure 1 that during infection, the spectral density of absorption capacity increases and the curve η_e(λ) shifts upwards in the range of 380–450 nm. In this case, the integral excitation parameter H, calculated by Formula (1), increases by 16.4%. There is a clear shift to the left, and the parameter H increases by 65.8% and 62.1% for the spectra of wheat and barley, respectively. The degree of infection of wheat and barley is 98%.

For the wavelengths of the excitation maxima λ_e equal to 232, 362, 424, 485 and 528 nm, the luminescence spectra φ_l(λ) were measured. Figure 2 shows the spectral characteristics of the luminescence of the oat seeds at different λ_e.

Figure 2 shows that all the spectra of infected seeds are located higher than those of healthy ones. The luminescence flux at λ_e = 232 nm Φ₂₃₂ increases significantly (by 1.71 times) during seed infection, but the luminescent signal is very low. When excited by λ_e = 424 nm, the photoluminescence flux increases by 1.30 times. For other excitation wavelengths, the flux increases significantly less—by 1.09–1.15 times. More detailed results for wheat are given in [27].

Additionally, the photoluminescence spectra of the wheat and barley seeds samples were measured at other degrees of β infection, obtained by mixing healthy seeds from the β = 0 sample and infected seeds from the β = 98% sample in appropriate proportions. The radiation fluxes calculated by Formula (2) are presented in

Table 1. Photoluminescence fluxes of plant seeds of various degrees of infection with fusarium.

Degree of Infection with Fusarium β, %	Φ232, r. u.	Φ362, r. u.	Φ424, r. u.	Φ485, r. u.	Φ528, r. u.
Wheat
0	36	387	833	501	222
25	56	490	823	502	240
50	62	507	762	514	243
75	71	572	729	524	244
98	95	556	746	545	264
Barley
0	21	152	653	694	298
25	28	232	638	606	301
50	42	243	678	595	280
75	45	304	683	592	268
98	66	386	870	533	257
Oats
0	21	586	1312	652	299
25	33	595	1517	691	319
50	35	643	1534	717	327
68	57	701	1572	726	334
82.4	36	641	1706	745	343

.

The dependences of photoluminescence fluxes on the degree of contamination of the sample of barley seeds Φ_λ (β) are shown in Figure 3.

4. Discussion

Based on the previously obtained results in the study of seed maturation [28], it can be assumed that infection with fusarium slows down the maturation of seeds. Therefore, in infected seeds, the excitation spectrum is closer to that of immature seeds. From the point of view of biochemistry, quantitative and qualitative changes in the spectra are associated with changes in the structure and chemical composition of the surface of the grain, namely the substitution of polysaccharides and proteins of cereals in the process of absorption and modification by mycoculture of Fusarium sporotrichioides fungi, as well as the effect on the spectra of pigments of this pathogen (fusarins and arofuzarins) and mycotoxins (DON, Zel, T2 and NT2).

The obtained dependences Φ_λ (β) are approximated by linear regression models with determination coefficients R² = 0.83–0.95, that is, they are statistically reliable. The exception is the dependence for barley at λ_e = 424 nm, where R² = 0.63.

Inverse dependences of the degree of contamination of barley seeds on their photoluminescence flux are obtained:

$$\mathsf{\beta }=2.17{\mathsf{\Phi }}_{232}-38.1,$$

(3)

$$\mathsf{\beta }=0.43{\mathsf{\Phi }}_{362}-64.9,$$

(4)

$$\mathsf{\beta }=0.33{\mathsf{\Phi }}_{424}-180.6,$$

(5)

$$\mathsf{\beta }=-0.62{\mathsf{\Phi }}_{485}+422.7,$$

(6)

$$\mathsf{\beta }=-1.97{\mathsf{\Phi }}_{528}+602.9.$$

(7)

For wheat seeds, they are the same:

$$\mathsf{\beta }=1.75{\mathsf{\Phi }}_{232}-62.6,$$

(8)

$$\mathsf{\beta }=0.49{\mathsf{\Phi }}_{362}-196.6,$$

(9)

$$\mathsf{\beta }=-0.76{\mathsf{\Phi }}_{424}+641.4,$$

(10)

$$\mathsf{\beta }=2.0{\mathsf{\Phi }}_{485}-1004,$$

(11)

$$\mathsf{\beta }=2.4{\mathsf{\Phi }}_{528}-537.6.$$

(12)

For oat seeds:

$$\mathsf{\beta }=1.80{\mathsf{\Phi }}_{232}-20.3,$$

(13)

$$\mathsf{\beta }=0.57{\mathsf{\Phi }}_{362}-314.1,$$

(14)

$$\mathsf{\beta }=0.22{\mathsf{\Phi }}_{424}-290.1,$$

(15)

$$\mathsf{\beta }=0.91{\mathsf{\Phi }}_{485}-595.8,$$

(16)

$$\mathsf{\beta }=1.95{\mathsf{\Phi }}_{528}-586.9.$$

(17)

The relative sensitivity of the flow change when the degree of infection changes should be determined by the formula:

$$S=100·\left|\frac{{\Delta \mathsf{\Phi }}_{\lambda }}{∆\mathsf{\beta }·{\mathsf{\Phi }}_{\mathsf{\beta }=0}}\right|.$$

(18)

The greatest sensitivity for barley is when luminescence is excited at the wavelength λ_e = 232 nm, which is 2.19 percent of the flow change per percentage of the contamination change. It is similar for wheat, where the maximum sensitivity at the wavelength λ_e = 232 nm is 1.59.

However, in practice, it is advisable to determine the degree of contamination by the ratio of flows Φ_λ1/Φ_λ2 which makes it possible to grade the measuring device in relative units. It will also help to increase the sensitivity if the wavelength λ₂ is chosen for the incident dependence β (Φ_λ). When calculating, we obtain that for barley, the highest sensitivity (3.13) is achieved with a flow ratio of Φ₂₃₂/Φ₄₈₅, and for wheat, the highest sensitivity (1.89) is achieved with a flow ratio of Φ₂₃₂/Φ₄₂₄. The grading equation for barley is

$$\mathsf{\beta }=1047\frac{{\mathsf{\Phi }}_{232}}{{\mathsf{\Phi }}_{485}}-23.0,$$

(19)

and for wheat:

$$\mathsf{\beta }=1219\frac{{\mathsf{\Phi }}_{232}}{{\mathsf{\Phi }}_{424}}-52.1.$$

(20)

The determination coefficients for Equations (19) and (20) are 0.92 and 0.98, respectively. The dependency graphs are shown in Figure 4.

For oat seeds, Equation (15) can be applied, the coefficient of determination for which is 0.88.

Based on the results obtained, a method of determining the degree of infection of seeds with fusariosis was developed. The block diagram of the method is shown in Figure 5.

The proposed method of determining the degree of seed contamination includes the following steps:

• Sample preparation is carried out, i.e., several seeds are taken from several plants of the same culture for research, and additional parameters are measured, for example, humidity. Then, they are placed in a dark, light-tight chamber.
• Photoluminescence is excited by two radiation sources (LEDs) in the short-wave 232 nm and medium-wave 424 nm/485 nm ranges.
• Luminescence is recorded by two photodetectors (photodiodes) with operating ranges: the first is 290–380 nm; the second is 450–600 nm or 520–650 nm.
• The electrical signal (the obtained ratio Φ₁/Φ₂) from photodiodes is amplified by an amplifier, converted into digital form and fed to the microcontroller.
• The ratio of photoluminescence fluxes and the degree of infection, taking into account a priori information, are calculated using the microcontroller.
• The obtained result is sent to the output indicator device. Taking into account the obtained data, the degree of seed infection is determined.

For oat seeds, it is possible to determine the degree of infection by Equation (15), but it is also possible to use the flow ratio, for example, Φ₄₂₄/Φ₄₈₅, to unify the methodology.

The method developed by the authors for diagnosing the infection of seeds with fusarium has limitations: seeds must be of conditioned humidity and have low contamination (no more than 5%).

5. Conclusions

The infection of grain seeds with fusarium affects the qualitative and quantitative characteristics of photoluminescence; the ratio between the maxima of the spectra and the integral parameters change. The method of determining the degree of infection can include both absolute measurements of the photoluminescence flux in the range of 290–380 nm and measurements of the flux ratios when excited by radiation of 232 nm and 424 nm for wheat and 485 nm for barley. The developed technique is fast-acting, inexpensive and easy to use, which does not require the pre-processing of samples.

Based on the methodology of determining the degree of infection of agricultural plant seeds, an optoelectronic device can be designed for the remote monitoring of the infection of their seeds. The proposed device can be assembled in the form of a nozzle on a UAV or a separate device.