Application of Digital Image Analysis for Assessment of Starch Content and Distribution in Potatoes

Abstract

This study presents the possibility of using digital image analysis for the assessment of the starch content and distribution in potatoes. Tubers of six cultivars that were stored for 3 months in contrasting conditions (4 °C vs. −15 °C) were used in the experiment. The starch distribution in the potato tubers was assessed on the basis of histograms of the pixel values along four lines in the tuber cross-sections. Next, the basic statistics were calculated and used for the analysis of variance. The applied method allowed more precise distinguishing between the studied potato cultivars than comparing the total starch content alone. The new method also clearly distinguished potatoes stored in a freezer from those kept in a cold store.

1. Introduction

Potatoes are one of the most popular crops in Europe and Asia. They are a tuberous plant with a high nutritional value and a low calorific value. Potato tubers are rich sources of energy and proteins, as well as vitamin C, folic acid, and group B vitamins [1,2]. One of the major technological parameters of potatoes is their starch content. It depends on the genetic features of the cultivars and, to a lesser extent, on tillage and crop husbandry. Starch is the major energy storage compound of plants, composed of two types of molecules: amylose and amylopectin. It is mostly used as a stabilizer, a thickener, and a fat substitute, and after modification it can serve as a substance improving texture and consistency [3,4,5].

The starch content of a potato can be studied using various methods, e.g., on the basis of the specific weights of the potato tubers or using a polarimetric method developed by Evers and modified by Grossfeld. The specific weights of potatoes depend on their starch contents, so in the former method their values are assessed with the use of a hydrostatic scale (according to Reimann-Parow) or the Krocker method [6,7,8,9]. In that method, the samples of tubers are soaked in a solution of sodium chloride and more sodium chloride is added until the density of the solution is equal to the mean specific weight of the tuber sample. Next, the solution is moved to a graduated cylinder, and its density is measured with a hydrometer. Finally, the starch content is read from a table on the basis of the measured density. The polarimetric method developed by Evers and modified by Grossfeld, Hadorn, and Bifer, consists of the removal of water-soluble substances from grated potatoes, the dissolving of the starch in a diluted HCl solution, clarification, and measuring the angle of rotation of the light polarization plane [10,11,12,13]. As described above, the determination of the total amount of starch in the potatoes is relatively easy to perform, however, no one has attempted to determine the distribution of starch in the potato tuber.

Another recent method for the assessment of potato starch content is digital image analysis, which is less time-consuming, less labor-intensive, and more accurate [14]. Currently, imaging systems are applied in many fields of research: medicine, geodesy and cartography, automation, forensic science, agriculture or the food industry, and many others [15,16]. Thanks to computer technology, data analysis is less expensive and tests are faster, repeatable, recorded automatically, and can be performed in difficult conditions which were inaccessible to humans in the past [15].

ImageJ software is a tool for image analysis used in many fields of research [16,17]. It was first released in 1997 by Wayne Rasband of the National Institutes of Health in the USA [14]. It is an open-source tool that runs on Linux, Mac OS X, and Windows, in both 32-bit and 64-bit modes. It has many functions typical of graphic software, but also some untypical ones, e.g., the designating of regions of interest (ROIs) in linear or other shapes and calculating the mean, standard deviation, and other statistics of the pixel values in the ROI [16]. Additionally, the software is easy to use for both amateurs and programmers, so it is now commonly used by scientists at various stages of their careers [17,18].

Like specialists in other fields of research, food technologists also attempt to simplify some procedures to save time, labor, and money. Because of the growing popularity of ImageJ software in other disciplines and the limited number of reports on the use of ImageJ software for starch content assessment in potatoes, we decided to verify the possibilities of its application for this purpose [19]. More precisely, this study aimed to determine the usefulness of ImageJ software for the assessment of the concentration and distribution of starch in potato tubers. Additionally, we checked the possibility of using this method to distinguish between the potato cultivars and to detect changes in the starch distribution in potatoes during storage at contrasting temperatures.

2. Materials and Methods

The experimental materials included 6 potato varieties (

Table 1. Potato varieties used in this study.

Variety	Culinary Type	Flesh Color	Origin
Allianz	A (boiling, very early)	white	Piaseczno, Poland
Gala	B (all-purpose, early)	pale yellow	Błaszki, Poland
Lilly	BC (all-purpose/baking, moderately early)	white	Piaseczno, Poland
Lord	AB (boiling/all-purpose, very early)	pale yellow	Piaseczno, Poland
Melody	BC (all-purpose/baking, moderately early)	pale yellow	Błaszki, Poland
Satina	B (all-purpose, moderately early)	yellow	Germany

).

Before the experiment, the potatoes were kept in a cold store at 4 °C or in a freezer at −15 °C for 3 months. Each sample was composed of 8 tubers of one variety. At the beginning of the experiment, each tuber was halved: one half was used for a starch content assessment with the polarimetric method developed by Evers and modified by Grossfeld, Hadorn, and Bifer, while the other half was subjected to a digital image analysis with ImageJ software. The polarimetric starch content assessment was repeated thrice for each tuber (3 replications) [20].

For the digital analysis, each half of a potato was cut into 6 slices of 3 mm thickness and next soaked in 0.5% KI for 1 min to stain the starch granules. Next, the slices were dried on paper and placed on the scanner bed for data acquisition and the digital recording of the image. The image acquisition was conducted for transmitted light: we used overhead lighting with two Diall LED bulbs placed 30 cm above the scanner bed and 15 cm apart from each other emitting neutral white light (4000 K, 1521 lm). Figure 1 shows the spectrum of light emitted by the bulb. The spectrum was measured using the AS7265x sensor.

For the digital image analysis, ImageJ software was used (https://imagej.nih.gov/ij/, accessed on 10 Dec 2022). Each scan of a potato slice was exported to ImageJ software and transformed into 8-bit greyscale. Next, the range of image brightness was corrected to increase the contrast between the stained and unstained parts of the potato. All the images of potato slices were corrected in an identical way, determined experimentally; the minimum and maximum values of brightness were set at 45 and 130, respectively (Figure 2).

After the image calibration, histograms of the pixel brightness (Figure 3) were determined along 4 lines crossing at an angle of 45° (Figure 4C). Next, the results were subjected to statistical analysis using STATISTICA 13 software [21]. The basic statistics of the histogram data were calculated: the mean, median, standard deviation, skewness, and coefficient of variation. Next, the basic statistics were used for a 2-way analysis of variance (ANOVA) to examine the influence of the storage conditions (cold store vs. freezer) and the potato cultivar. Homogeneous groups were determined by a Duncan test at a significance level of α = 0.05.

3. Results and Discussion

The results of the ANOVA of the starch content in relation to the potato variety (

Table 2. Starch contents of cold-stored potatoes of 6 cultivars assessed chemically (homogeneous groups are labeled with the letters).

Potato Cultivar	Starch Content [%]
Gala	11.7 a
Lord	12.5 a
Melody	12.6 a
Lilly	14.2 b
Satina	17.7 c
Allianz	17.9 c
LSD value	1.09

) allowed us to distinguish three homogeneous groups: potatoes with a low (Gala, Lord, Melody), medium (Lilly), and high starch content (Satina and Allianz). The starch content assessment with the polarimetric method did not enable us to distinguish between individual varieties.

The starch contents significantly differ between the frozen and the cold-stored potatoes (

Table 3. Starch contents of frozen and cold-stored potatoes, assessed chemically (homogeneous groups are labeled with the letters).

Storage	Starch Content [%]
Cold store: 4 °C	14.9 a
Freezing: −15 °C	14.0 b
LSD value	0.64

), but the differences are smaller than the differences between the cultivars, so it does not make it possible to unambiguously distinguish the potatoes stored in a freezer. The decrease in the starch contents of frozen potatoes is linked with the starch degradation to simpler sugars and is the plant’s protective mechanism in response to freezing [22,23]. At the same time, water is detached from the starch at low temperatures, leading to its retrogradation [24,25,26].

The results of the ANOVA of the basic statistics of pixel brightness in relation to potato variety (

Table 4. Values of basic statistics of pixel distribution along the measurement lines in cold-stored potatoes, depending on potato cultivar (homogeneous groups are labeled with the letters).

Potato Cultivar	Mean	Median	Range	Skewness	Coefficient of Variation
Gala	68.2 a	68.0 a	53.7 a	0.29 abc	42.4 a
Lord	62.5 ab	62.1 ac	51.9 a	0.36 c	28.4 b
Melody	54.8 b	54.4 c	59.8 a	0.18 d	19.0 c
Lilly	58.1 b	57.6 c	63.2 a	0.26 ab	25.1 d
Satina	44.3 c	43.6 b	61.3 a	0.32 bc	26.4 e
Allianz	43.9 c	43.0 b	61.7 a	0.21 ad	26.9 e
LSD value	8.7	8.3	17.4	0.07	1.1

), allowed us to distinguish only three groups of cultivars, very much like in the case of the chemical analysis of the starch content.

The mean values and medians are significantly negatively correlated with the total starch contents of the potatoes because the values similar to 255 denote light-colored areas, which are not stained with iodine (lack of starch), while values close to 0 denote dark areas, i.e., starch granules stained with iodine. The starch contents of the potatoes vary from 10% to 20% and depend on the potato variety, but they do not enable precise distinguishing between the cultivars as in the work of Jakab et al., 2014, which compared the starch contents of different apple varieties using ImageJ [27].

The analysis of the range values did not allow the distinguishing of any homogeneous groups. This is not surprising, as it means that the potato tubers contain starch-rich areas and areas devoid of starch. The range of pixel brightness in all the samples is nearly identical and varies from 51.9 to 63.2.

An analysis of skewness shows that the starch distribution in potatoes is right-skewed and skewness values are positive and range from 0.18 for cv. Melody to 0.36 for cv. Lord. This coefficient enabled us to distinguish four homogeneous but overlapping groups, so it cannot be used to distinguish between the potato cultivars.

The most interesting results concern the coefficient of variation. This value is associated with the pattern of starch distribution along the measurement lines, irrespective of the total starch content. The analysis of this coefficient allows the distinguishing of all the studied cultivars except Satina and Allianz, which formed a homogeneous group. This suggests that the patterns of starch distribution in potato tubers are cultivar-specific. Figure 3 shows some light-colored areas in the tuber interior, where less starch is accumulated than under the skin. The shape of the light-colored area is cultivar-specific. A similar study was performed by Jakab et al. using images taken with ImageJ software [27]. We also investigated the correlation between the starch content assessed chemically and the mean pixel brightness (Figure 5). The coefficient of correlation was high and reached −0.93, and the formula for estimating the starch contents of potatoes on the basis of the mean pixel brightness is:

$$S=28.84-0.26·Px$$

(1)

where:

$S$ = starch content in potato [%]

Px = mean pixel brightness determined by digital image analysis.

Figure 5. Dependence of starch contents of potatoes on mean pixel brightness assessed using digital image analysis, R² = 0.86.

Figure 5. Dependence of starch contents of potatoes on mean pixel brightness assessed using digital image analysis, R² = 0.86.

An ANOVA was performed for the basic statistics of pixel distribution along the measurement lines, depending on the potato storage conditions (

Table 5. Values of basic statistics of pixel distribution along the measurement lines, depending on potato storage conditions (homogeneous groups are labeled with the letters).

Storage	Mean	Median	Range	Skewness	Coefficient of Variation
Cold store: 4 °C	57.9 a	57.3 a	58.5 a	0.24 a	17.4 a
Freezing: −15 °C	98.7 b	92.3 b	202.7 b	0.29 b	38.6 b
LSD value	5.0	5.2	10.0	0.04	0.6

). All the analyzed coefficients significantly differed between the two groups. The very high mean, median, and range values allow the distinguishing of frozen potatoes from cold-stored potatoes. The storage of potatoes in a freezer results in a significant decrease in the starch contents of the tubers, so the mean values and medians of the starch contents of the potatoes allowed us to distinguish two homogeneous groups. It is noteworthy that the freezing of potatoes does not cause a uniform transformation of all the starch deposits into simpler sugars [22], but influences the coefficient of variation of the starch granule distribution along the measurement lines. This means that in some parts of the tubers, starch is decomposed faster to simpler sugars than in other areas.

4. Conclusions

The use of digital image analysis facilitates the assessment of both the starch content and distribution in potatoes. The high coefficient of correlation between the starch contents assessed chemically and the mean pixel brightness determined by digital image analysis allows its accurate estimation with the latter method. The tested method enabled us to distinguish between five of the six potato varieties used in this study. It is noteworthy that the chemical method for the starch content assessment in potatoes allowed us to distinguish only three groups of cultivars. Interestingly, the pattern of starch distribution in the potatoes is cultivar-specific and makes it possible to distinguish between the varieties better than only on the basis of the total starch content. Digital image analysis can be used for fast estimation if the material delivered to a potato-processing plant is homogeneous in terms of potato variety and if the potatoes were subject to long-term freezing or not. Simultaneously, the assessment of the starch content and distribution in potatoes by using digital image analysis is markedly cheaper and less time-consuming than chemical methods.