Introduction

Jiló (Solanum gilo Raddi) (Solanaceae) is native from India and was introduced in Brazil by slaves. Its fruits, which are edible, are light green or dark green color, becoming reddish orange when ripe (^{Filgueira, 2003}). Jiló, also known as Scarlet eggplant or Brazilian red eggplant, is a vegetable with good acceptance in the market, especially in the Southeast region of Brazil. However, there are few studies with this species, mostly addressed to seeds (^{Alves et al., 2012}).

The maturation degree of seeds influences their quality, and immature seeds have low vigor and germination power (^{Carvalho et al., 2009}). The period from anthesis to the physiological maturation varies from one species to another, and sometimes from one cultivar to another (^{Araújo et al., 1982}). For some species with fleshy fruits, seeds will continue their maturation after harvest (^{Barbedo et al., 1994}; ^{Justino et al., 2015}). This is an advantage because it decreases the number of harvests and enables to harvest fruits with different ripeness stages (^{Mengarda and Lopes, 2012}).

Defining the physiological maturation of seeds is important to estimate time of harvest. Thus, it is necessary to evaluate physiological quality of seeds and these evaluations are most commonly carried out through germination and vigor tests. However, X-ray analysis is a promising technique for seed quality evaluation (^{Gomes Júnior et al., 2012}), helping to estimate harvest time.

^{Simak and Gustafsson (1953)} carried out the first evaluation of seeds with conifer species by using X-ray analysis. This technique analyzes the internal structure of the seed through its exposure to a source of low X-ray energy; the radiation fractions enter the seed and reach the film, allowing formation of an image characterized by different shades of gray. The principle of the technique is the absorption of X-rays in different amounts by the seed tissues due to its structure, composition and density, in addition to radiation exposure period (^{ISTA, 2004}).

Image analysis to evaluate the physiological quality of seeds can help reducing the subjectivity of analysis, as it eliminates human error, making it a quicker and more efficient process. Image analysis of seedlings and seeds is outstanding in this regard for several species (^{Nunes et al., 2014}), but for Solanaceae seeds, such as jiló, there is still little information.

The results of the X-ray test have successfully related internal morphology of seeds with germination or morphology of tomato (Lycopersicon esculentum) (^{Van der Burg et al., 1994}), pepper (Capsicum annuum) (^{Gagliardi and Marcos-Filho, 2011}; ^{Dell’Aquila, 2007}), eggplant (Solanum melongena) (^{Silva et al., 2012}), Acca sellowiana (^{Silva et al., 2013}) and pumpkin (Cucurbita pepo) seedlings (^{Silva et al., 2014}). However, in most cases, this classification is visually performed; therefore, in order to develop consistent assessment models to define extension categories of embryonic development or free space inside the seeds, there is a need for more accurate methods (^{Marcos-Filho et al., 2010}).

There are no studies for S. gilo that relate the internal morphology of the seeds with their maturation stage through X-ray image analysis. Thus, the aim of this study was to relate the internal morphology of jiló seeds at different stages of maturation, with germination, using radiographic images.

Materials and Methods

The jiló seed production was carried out between July/2014 and January/2015 at Palmital farm in Ijaci, MG, Brazil (21° 9' 24" S, 44° 55' 34" W). This farm has an average altitude of 918 m and its soil is classified as clayey Oxisol.

In the first stage of the study, jiló seedlings came from Morro redondo (dark green round - DGR) and Tinguá (light green long - LGL) varieties. Seeds of both varieties of jiló were sown in plastic trays (polystyrene) of 128 cells, with Tropstrato HA - Vegetables commercial substrate, and then covered with vermiculite. Forty days after sowing, seedlings were transplanted to a greenhouse and planted in 38 m long beds with 21 plants each, spaced 1 m between rows and 0.60 m between plants.

Flowers were daily labeled on the day of anthesis to obtain the number of fruits to guarantee enough amounts of seeds for all analyses. Treatments for each variety were: T1) fruits harvested at 35 d after anthesis (DAA); T2) 40 DAA; T3) 45 DAA and, T4) 45 DAA + 7 d. In T4, fruits harvested at 45 DAA remained at rest for 7 d in a dark and ventilated space in the seed laboratory (LAS). The experimental design was completely randomized with four replicates of 50 seeds from each treatment.

Seeds were manually extracted and washed in running water. After extraction, seeds were arranged in screens for 24 h in a laboratory environment and then treated to slow drying in an air circulation oven (Nova Ética - 420) at 35 °C until reaching about 8 % water content (^{Queiroz et al., 2011}). These seeds were X-rayed and subsequently submitted to germination test. In the radiographic analyzes four replicates of 50 seeds were placed on a transparent adhesive tape (double-sided) and fixed in transparent plastic foil. Seeds from each treatment were numbered according to the position occupied on the blade. The plastic foil was placed inside the digital X-ray equipment, Faxitron^® brand MX-20 DC-12 Model, and subject to radiation for 12 s at 26 kV.

After removal from the transparent film, seeds were transferred to acrylic boxes (gerbox type) (110 x 110 x 35mm) following the same order they were in the X-ray images. Afterwards, the germination test was carried out on two sheets of blotting paper moistened with distilled water in the proportion equivalent to 2.5 times the paper mass. Seeds were distributed in acrylic boxes (gerbox type) in the same positions they were in the radiographic images and kept in seed germinator chamber (B.O.D. type) under alternate temperature and light, 20 °C / 16 h dark, 30 °C / 8 h with light, according to RAS (^{BRASIL, 2009}).

The count was made 14 d after sowing (^{BRASIL, 2009}). Normal seedlings (NS), abnormal (AS) and dead seeds (DS) were photographed with a Canon SX50^® digital camera. The images of X-rayed seeds were saved in JPEG format and analyzed with the ImageJ software, adapting the analysis methodology of analysis used in other studies to leaf area measurement and seed internal area of seeds (^{Silva et al., 2013}).

The Analysis steps in ImageJ program Version 1.50i were as follows: image opening and its conversion to grayscale type in 8 bits followed by a selection of interest area of interest for analysis; scale selection for image calibration, and in this study the amount in mm of each image (231 x 210 mm) was considered as a reference. Variables were: the object area, the perimeter, the standard deviation, and the fraction area, which represents the internal free areas of seeds. Color adjustment was made in order to separate the areas of interest from other image constituents; and finally, measurement of variables thus obtaining the results. All procedures were manually performed for each seed.

The free space in the internal seed cavity was calculated using the ImageJ software in four replicates of 50 seeds for each treatment. In order to prevent biases by possible damages in seeds due to manual extraction, we evaluated the averages of the free space in the internal cavity of the seed, germination, normal and abnormal seedlings and dead seeds of each treatment. Data from the germination test and the seeds image analysis were used for ANOVA analysis, and treatments means were compared with the Scott-Knott test (p ≤ 0.05).

Results and Discussion

Tones in the radiographic image analysis are defined by the level of radiation absorption in different areas of the seed, which is determined by thickness, density and composition of tissues in the seed (^{Simak, 1991}; ^{ISTA, 1993}). Thus, seeds lacking embryonic tissue provide dark images because they do not have resistance to X-rays.

Radiographic images show the inner area of jiló with different percentages of free areas between the seed coat and endosperm - embryo at different maturation stages. All treatments for both varieties had free areas within the seeds. Figure 1 shows seeds of Morro Redondo and Tinguá varieties, from fruits harvested at T1 (35 DAA), T2 (40 DAA), T3 (45 DAA) and T4 (45 DAA + 7 d).

Figure 1 Radiographic images of jiló seeds (Solanum gilo), Morro Grande (A-D) and Tinguá (A’-D’) varieties at different stages of maturation. Dead seeds in T1 - 35 DAA (A, A’), Abnormal seedlings in T2 - 40 DAA (B, B’), and Normal seedlings in T3 - 45 DAA (C, C’), and T4 - 45 DAA + 7 d. (D, D’). Percentage represents internal free area. 

Other studies also verified the efficiency of X-ray image analysis for assessing the internal area of bell pepper seeds (^{Gagliardi and Marcos-Filho, 2011}), eggplant (^{Silva et al., 2012}), melon (Cucumis melo), pumpkin and watermelon (Citrullus lanatus) (^{Gomes Júnior et al., 2012}), and pumpkin (^{Silva et al., 2014}).

The ImageJ software allowed the measurement of internal free areas of seeds, which allowed evaluation of the relationship between the free inner area of seeds and germination. This result is interesting because the recommended software approaches for image analysis are: Tomato Analyzer^® (^{Marcos-Filho et al., 2010}) and Image ProPlus^® (^{Dell’Aquila, 2007}; ^{Silva et al., 2012}). However, we did not find image analysis studies with jiló seeds using the ImageJ software in the literature reviewed; there is only one study with Acca sellowiana seeds by ^{Silva et al. (2013)} using this software.

Therefore, the promising results of this study showed an alternative to explore different types of seeds, since computerized methods using high speed capture and image processing, are the most non-invasive advanced technique providing high efficiency level for analyzing quality of seeds (^{Dell’Aquila, 2009}).

Regarding the fruit harvest season (days after anthesis), there was no statistically significant difference in each stage of maturation between Morro Grande and Tinguá varieties, which have different colors and shapes, dark green round and light green long, respectively. However, there was a statistically significance difference in the free areas of seeds of both varieties (Table 1). T1 fruits of Morro Redondo and Tinguá varieties harvested at 35 DAA had a larger free average area of 21 and 23 %, respectively, as compared to fruits of T2 harvested at 40 DAA with a free average area of 17 and 18 %. T3 fruits harvested at 45 DAA had free average area of 10 and 9 %, and T4 fruits harvested at 45 DAA and rested 7 d showed free average area of 5 and 4 %, for Morro Grande and Tinguá varieties, respectively (Table 1 and Figure 1). Thus, there was a reduction in the free area as seed maturity advanced in both varieties evaluated. Figures 2 and 3 show the distribution of free areas of seeds of Morro Redondo and Tinguá varieties (n = 200) for all treatments.

Figure 2 Internal free area (%) occupied by embryo and endosperm of each jiló seed (Solanum gilo) in a sample of 200 seeds, treatment - Morro Redondo. T1 - 35 DAA (A), T2 - 40 DAA (B), T3 - 45 DAA (C), and T4 - 45 DAA + 7 d (D). 

Figure 3 Internal free area (%) occupied by embryo and endosperm of each jiló seed (Solanum gilo) in a sample of 200 seeds, treatment - Tinguá. T1 - 35 DAA (A), T2 - 40 DAA (B), T3 - 45 DAA (C), and T4 - 45 DAA + 7 d (D).  

According to ^{Carvalho and Nakagawa (2000)}, the further away from anthesis the seed is removed from the fruits, the heavier the seed is until reaching physiological maturity. ^{Miranda et al. (1992)} studied eggplant fruit maturation and concluded that there was no difference in dry weight gain between seeds collected at 50 DAA and in seeds harvested at 60 DAA. Paprika seeds at 55 DAA reached full physiological maturity, which is the point where the seeds reached the highest weight, and therefore, the best harvest time (^{Oliveira et al., 1999}). Similarly, for Morro Grande e Tinguá varieties of jiló seeds, with the advancement in the maturation stage, a greater reserve accumulation occurred, which was evidenced by the increase in the evaluated seed weight (^{Alves et al., 2017}).

Evaluation of seed internal morphology is important to characterize some understudied species, in order to improve quality of seed lots, regarding their physical and physiological attributes, since defective or empty seeds affects the germination results (^{Gomes Júnior, 2010}). Thus, the analyses carried out in our study are useful to separate high and low quality seed lots, contributing to the jiló seeds quality control.

At the end of the germination test (14 d) we counted the percentage of normal and abnormal seedlings and dead seeds of Morro Redondo and Tinguá varieties. Results for both Morro Redondo and Tinguá varieties were: T1 had no germination and 100 % percentage of dead seeds for both varieties; T2 had a germination percentage of 10 and 9 %, without normal seedlings, 20 and 17 % of abnormal seedlings, and 80 and 83 % of dead seeds, respectively; T3 had germination rates and percentage of normal seedlings of 86 and 84 %, there were no abnormal seedlings and resulted in a 14 and 16 % of dead seeds, respectively; T4 had 100 % normal seedlings for both varieties. There was no statistically significant difference between Morro Grande and Tinguá varieties for these characteristics. However, seeds from the T4 were statistically superior to T3, which was higher than T2 and T1 in both varieties for germination percentage (Figure 1 and Table 1).

There was a relation between germination and internal free area because, according to the moment of harvest (days after anthesis), the results were statistically higher from T1 to T4, meaning that seeds had a reduction in the empty area and an increase in germination percentage directly proportional to the maturity stage, as observed in Figure 1 and Table 1. Thus, harvest at 45DAA + 7 is recommended, since seed germination was 100 % and resulted in 100 % normal seedlings in both varieties. Similar results for dry weight gain, that is, free area reduction and increased germination from fruits harvested at different maturation stages and days after anthesis, were found by ^{Oliveira et al. (1999)}, ^{Carvalho and Nakagawa (2000)}, ^{Nakada et al. (2011)}, and ^{Santos et al. (2015)}.

It is noteworthy that the absence of normal seedlings for T1 and T2 can be directly related to the immaturity of the seeds and the high percentage of internal free area of the seed, as showed in the of X-ray images evaluation along with the program ImageJ. Furthermore, it is related to the physiological potential since, according to ^{Sediyama et al. (1991)}, early harvest can result in reduced seed germination and increased number of immature seeds.

In a similar study, ^{Dell’Aquila (2007)} found that in paprika seeds, when the free space between the embryo and endosperm was higher than 2.7 %, that is, seeds with endosperm area and less than 97.3% embryo, there was a reduction of germination with increased percentage of abnormal and non-germinated seedlings.

In studies with Cecropia pachystachya (^{Pupim et al., 2008}), aroeira (Schinus terebinthifolius) (^{Machado and Cicero, 2003}), paprika (^{Dell’Aquila, 2007}), eggplant (^{Silva et al., 2012}), watermelon (^{Gomes Júnior et al., 2012}) and pumpkin (^{Silva et al., 2014}), partially formed seeds with internal free area were unable to produce normal seedlings in the germination test. However, for many species of fleshy fruits, fruit harvest followed by a rest period of 7 or 10 d positively interfere in the seed quality (^{Dias and Nascimento, 2009}).

Thus, radiographic image analysis with the ImageJ software appears as a good alternative for the identification of internal free areas of jiló seeds (Solanum gilo), assisting to a greater or lesser physiological potential of lots in different maturation stages.

Conclusions

The radiographic image analysis for jiló seeds (Solanum gilo) using ImageJ software allows measurement of internal empty areas at different stages of maturation, as well as the determination of the relationship between emptiness degree and germination. Thus, it is possible to correlate these data with other attributes of seed quality. Moreover, the use of image analysis systems, as reported in this study, may contribute to improving the quality of seed lots, allowing the identification of seeds with greater potential to germinate, assisting sellers in their quality control of jiló, and helping to reduce economic losses.