Impaginato 165 Adv. Hort. Sci., 2025 39(3): 165­173 DOI: 10.36253/ahsc­17532 https://oaj.fupress.net/index.php/ahs Digital and multivariate analysis of lettuce seed vigor: Impact of hydropriming on physiological potential H.A. Trujillo 1 (*), F.G. Assis de Oliveira 2, C.M. Villegas Lobos 2, M. da Silva 2, F.G. Gomes­Junior 1 1 Department of Crop Science, University of São Paulo, ‘Luiz de Queiroz’ College of Agriculture, Av. 11 Pádua Dias, 13418‐900 Piracicaba, SP, Brazil. 2 Department of Exact Sciences, University of São Paulo, ‘Luiz de Queiroz’ College of Agriculture, Av. 11 Pádua Dias, 13418‐900 Piracicaba, SP, Brazil. Key words: Applied statistics, image analysis, physiological variability, seed priming. Abstract: Digital image analysis has emerged as a highly precise and efficient methodology for assessing the physiological attributes of seeds. This research aimed to assess the morphological and physiological properties of lettuce seeds subjected to hydropriming using multivariate statistical approaches. Two lettuce genotypes, Roxa and Vanda, were evaluated under hydropriming treatments (primed­dry and primed­stored). Seedlings were digitally scanned, and vigor indices were quantified using the Seed Vigor Imaging System (SVIS®). Data were analyzed by multivariate analysis of variance (MANOVA), with tests of normality and homogeneity of covariance ensuring analytical robustness. The primed­dry treatment resulted in minimal improvement in vigor and uniformity, while the primed­stored treatment promoted a partial recovery of these attributes. The Roxa genotype exhibited greater variability in vigor and seedling length, whereas Vanda demonstrated higher uniformity but slightly reduced seedling growth. A strong positive correlation was observed between the vigor index and seedling length, reinforcing the importance of these parameters in seed quality assessment. These findings underscore the utility of digital image analysis combined with multivariate statistical methods for the accurate assessment of seed vigor, thereby improving seed­lot classification and informing decision­making in lettuce production systems. 1. Introduction Computerized analysis of seed and seedling images has emerged as an (*) Corresponding author: heiberandrestrujillo@gmail.com Citation: TRUJILLO H.A., ASSIS DE OLIVEIRA C.F., VILLEGAS LOBOS C.M., DA SILVA M., GOMES­JUNIOR F.G., 2025 ­ Digital and multivariate analysis of lettuce seed vigor: Impact of hydropriming on physiological potential. ­ Adv. Hort. Sci., 39(3): 165­173. ORCID: THA: 0000­0001­6604­9438 AOCF: 0000­0003­2143­9015 VLCM: 0000­0003­3176­5236 DSM: 0000­0001­9079­8981 GJFG: 0000­0001­9620­6270 Copyright: © 2025 Trujillo H.A., Assis de Oliveira C.F., Villegas Lobos C.M., da Silva M., Gomes­Junior F.G. This is an open access, peer reviewed article published by Firenze University Press (https://www.fupress.com) and distributed, except where otherwise noted, under the terms of CC BY 4.0 License for content and CC0 1.0 Universal for metadata. Data Availability Statement: All relevant data are within the paper and its Supporting Information files. Competing Interests: The authors declare no conflict of interests. Received for publication 27 March 2025 Accepted for publication 19 August 2025 AHS Advances in Horticultural Science AHS ­ Firenze University Press ISSN 1592­1573 (on line) ­ 0394­6169 (print) http://doi.org/10.36253/ahsc-17532 http://oaj.fupress.net/index.php/ahs http://orcid.org/0000-0001-6604-9438 http://orcid.org/0000-0003-2143-9015 http://orcid.org/0000-0003-3176-5236 http://orcid.org/0000-0001-9079-8981 http://orcid.org/0000-0001-9620-6270 http://www.fupress.com http://creativecommons.org/licenses/by/4.0/legalcode http://creativecommons.org/publicdomain/zero/1.0/legalcode Adv. Hort. Sci., 2025 39(3): 165­173 166 innovative tool for determining the physiological characteristics of these structures. This technology stands out for its objectivity, specificity in detecting subtle traits, efficiency, and potential for standardization (Rahman and Cho, 2016; Xia et al., 2019; Wang et al., 2021; Liu et al., 2023). The use of digital imaging and automated software enables the analysis of a large number of samples in shorter periods, increasing the efficiency and accuracy of evaluations. Assessing the physiological potential of lettuce seeds through digital image analysis has proven to be a promising approach, particularly as a complement to conventional methods that do not fully reflect seed quality under real field conditions (Waters­ Junior and Blanchette, 1983; Marcos­Filho, 1999). The use of the Seed Vigor Imaging System (SVIS®), developed by Sako et al. (2001), has been widely adopted to quantify seed vigor in various species, including soybean, corn, melon, sweet corn, castor bean, peanut, okra, common bean, eggplant, tomato, cotton, and sunflower (Hoffmaster et al., 2003; Marcos­Filho et al., 2006; Marchi et al., 2011; Alvarenga et al., 2013; Caldeira et al., 2014; Gomes­ Junior et al. , 2014; Rocha et al. , 2015). This technology allows for detailed analyses of parameters such as seedling growth uniformity and development, reducing the subjectivity of traditional evaluations. However, to enhance the accuracy of vigor assessment, it is essential to employ multivariate statistical models that enable the simultaneous analysis of multiple interrelated variables. Multivariate Analysis of Variance (MANOVA) has been used to investigate complex interactions between experimental factors, allowing for the identification of patterns that would be difficult to detect using univariate approaches (Johnson and Wichern, 2002; Nicacio et al., 2013). Previous studies have demonstrated that MANOVA is an effective tool for evaluating seed performance under different treatments, ensuring greater robustness in result interpretation (Oliveira et al., 2013). This study aimed to analyze the morphological and physiological properties of lettuce seeds subjected to hydropriming using digital imaging of seedlings and a multivariate approach. The study sought to understand the interactions between the physiological attributes of the seeds and the impact of hydropriming on germination potential and early seedling development. 2. Materials and Methods The research was conducted at the Seed Analysis Laboratories, the ‘Professor Silvio Moure Cicero’ Image Analysis Laboratory of the Department of Crop Science, and the Department of Math, Chemistry, and Statistics at the Luiz de Queiroz College of Agriculture, University of São Paulo in Piracicaba, SP, Brazil. Seed material and priming treatment Lettuce seeds from the genotypes Scarlet Red Crisphead (Roxa) and Vanda Crisphead (Vanda) were used, supplied by Sakata Seed South America Ltd. The selection of the two lettuce genotypes was based on their contrasting physiological and morphological characteristics and on their commercial relevance within Brazilian lettuce production systems. Each genotype was represented by ten seed lots, with germination rates within commercial standards and different vigor levels among the lots. Each seed lot was divided into three treatments: (i) non­primed seeds (control), (ii) hydroprimed dried seeds (dried in an oven at 30°C and 45­55% relative humidity for 96 hours), and (iii) hydroprimed stored seeds (dried and stored in a chamber at 10°C and 30% relative humidity for three months). Hydropriming was performed using the drum method, utilizing the S­HIDRO® Control equipment, which allowed for the controlled application of water at regular intervals until reaching the required volume for each lot (Kikuti and Marcos­Filho, 2012). The calculation of the required water volume for this method was based on the water imbibition curve, considering the volume needed for each seed lot before primary root protrusion (Caseiro, 2003). At the beginning of each cycle, the electric pump was activated for 1 second, allowing the intake of a water volume between 0.9 and 1.1 ml, adjusted according to the specific needs of each seed lot and accounting for system losses to ensure 100% efficiency. Water application was carried out at one­hour intervals until the total required volume for each seed lot was reached (ranging from 4.00 to 4.31 ml for the Roxa genotype and from 3.67 to 3.87 ml for Vanda). The entire procedure was conducted under laboratory conditions at a constant temperature of 25°C. Seed vigor assessment using digital image analysis (SVIS® Software) Four replicates of 25 seeds per lot for each Trujillo et al. ‐ Multivariate and digital analysis of lettuce seed vigor 167 Vigor results from the combination of two main components: average seedling length and uniformity, both adjusted by weighting factors (W) defined by the system, allowing the assignment of greater or lesser relative importance to each characteristic (hypocotyl­to­radicle ratio of 40:60, applied in this research). Growth is estimated from the mean lengths of the hypocotyl (lh) and radicle (lr), weighted by their respective coefficients (Wh and Wr), thereby composing the total seedling length. Uniformity, in turn, is calculated based on the standard deviations of hypocotyl length (Sh), radicle length (Sr), total seedling length (Stotal), and the hypocotyl­to­radicle length ratio (S(r/h)), also weighted by their respective coefficients (Wsh, Wsr, W(sr/h)). Multivariate analysis Multivariate Analysis of Variance (MANOVA) was used to describe the effects of categorical factors (treatments) on multiple response variables (Huberty and Olejnik, 2006). MANOVA extends ANOVA to the multivariate context, allowing simultaneous testing of multiple dependent variables. In this study, a two­ way MANOVA was employed, considering two categorical factors: genotype (Roxa and Vanda) and hydropriming treatment (control, primed dry, primed stored). This factorial approach allows for the evaluation of both the main effects of each factor and their interaction. The general model for Two­Way MANOVA: Xlkr= μ + τl + βk + γlk + ϵlkr X represents the dependent variable, l represents the levels of factor 1 (genotype); k represents the levels of factor 2 (hydropriming treatment); r represents the replications; µ is the overall mean; τl representing the interaction effect and βk representing the treatment or genotype main effects; γ lk is the interaction effect between the factors; represents the random error. The hypothesis tests included: Interaction effect: H0 : γ11 = γ12=⋯= γgb = 0 vs. H1 : at least one γgb ≠ 0 Genotype effect: H0 : τ11 = τ12=⋯= τg = 0 vs. H1 : at least one τg ≠ 0 Hydropriming effect: H0:β1 =β2 =β3 = 0 vs. H1 : at least one βb ≠ 0 Statistical significance was determined using Wilks’ lambda (Wilks, 1935), Pillai’s trace (Hand and Taylor, 1987), Hotelling­Lawley trace (Krzanowsk and treatment were arranged in two rows on the upper third of two blotter paper sheets, placed on the lids of transparent plastic boxes (11 × 11 × 3.5 cm). The boxes were covered with transparent plastic bags and incubated in a BOD chamber at 25°C for three days in darkness. To ensure proper seedling development according to natural geotropism, the boxes were positioned at a 70° angle relative to the horizontal plane. To determine seed vigor, the Seed Vigor Imaging System (SVIS®) (Sako et al., 2001) software was used. The seedlings (and ungerminated seeds) from each replicate were transferred onto a blue ethylene­vinyl acetate (EVA) sheet, providing the necessary contrast for system analysis. The seedlings were then scanned using an HP Scanjet 200 scanner, which was inverted and placed inside an aluminum box (60 × 50 × 12 cm), with the resolution set to 300 dpi and connected to a computer. The scanned images were processed using SVIS® software, including manual corrections when necessary to ensure accurate seedling identification (Fig. 1). The seed vigor index is calculated according to the methodology proposed by Sako et al. (2001): Vigor index = WG × Growth × Wu × Uniformity Seedling length = WG {Wh × lh + Wr × lr, 1000} Uniformity index = max {1000‐(Wsh × Sh + Wsr × Sr + Stotal + W(sr/h) × S (r/h)‐Wd ),0} Fig. 1 ­ Workflow of computerized image analysis of lettuce seedlings, from germination to vigor assessment. Steps include seedling acquisition from the germination test (A), transfer to an EVA sheet (B), digital imaging (C), and vigor determination using the SVIS® software (D). The results include the vigor index (VI) and the uniformity index (UI), both ranging from 0 to 1000 (directly propor­ tional to seedling vigor), as well as the average seedling length, initially measured in pixels and later converted to centimeters. Adv. Hort. Sci., 2025 39(3): 165­173 168 Marriott, 1994; Anderson, 2003), and Roy’s largest root (Krzanowski, 2000). When MANOVA indicated significant differences, post hoc univariate ANOVAs were conducted to determine which dependent variables contributed to the observed differences. Statistical assumptions and data validation Before conducting MANOVA, the following statistical assumptions were tested: multivariate normality, using the Henze­Zirkler test (Henze and Zirkler, 1990) and homogeneity of covariance matrices using Box’s M test (Johnson and Wichern, 2002). In the univariate context, the Anderson­ Darling test (Scholz and Stephens, 1987) was used to assess normality. Initial analyses revealed that seedling size did not meet the assumption of normality; therefore, this variable was removed from the final MANOVA model to ensure compliance with statistical assumptions. Software and data processing All statistical analyses were performed using the R programming language (R Core Team, 2024), with the packages ‘MVN’ for normality tests and ‘car’ for MANOVA. Data visualizations, including boxplots and correlation matrices, were generated using the ‘ggplot2’ and ‘corrplot’ packages. Image processing was performed using SVIS® software, ensuring standardization and reproducibility of seed vigor measurements. 3. Results Descriptive data analysis The descriptive analysis allowed the identification of the main characteristics of the studied variables. Figure 2 presents boxplots for the three response variables in this research: vigor index, uniformity index, and seedling length. It is observed that the means values of vigor and uniformity are similar between genotypes, while the dispersion of vigor is higher. Seedling length, measured in centimeters, is on a different scale from the other variables and exhibits a lower correlation with them. The Roxa genotype exhibits a higher median and greater variability, suggesting either greater vigor or increased heterogeneity. In contrast, the distribution of the uniformity index between the two genotypes is similar, indicating that growth is uniformly distributed. The boxplots for the analyzed variables concerning hydropriming treatments are shown in Figure 3. It is observed that vigor and uniformity vary significantly between treatments. Seeds subjected to the primed dry treatment showed lower means for these variables. Seedling length showed less pronounced differences, with the control treatment presenting the highest mean. The mean values of the variables for each genotype are shown in Table 1, while Table 2 shows the corresponding values for each hydropriming treatment. Table 1 ­ Mean values of vigor index, uniformity index, and seedling length for Roxa and Vanda genotypes Fig. 2 ­ Boxplot comparing the vigor index, uniformity index (both ranging from 0 to 1000), and seedling length (cm) between Roxa and Vanda genotypes. Fig. 3 ­ Boxplot comparing the vigor index, uniformity index (both ranging from 0 to 1000), and seedling length (cm) across control, primed dry, and primed stored treatments. Genotype Vigor index Uniformity index Seedling length (cm) Roxa 761 762 3.47 Vanda 754 757 3.38 Trujillo et al. ‐ Multivariate and digital analysis of lettuce seed vigor 169 The distribution, correlation, and dispersion of the vigor, uniformity, and seedling length, their correlations, and dispersion are shown in figure 4. The vigor index is strongly correlated with seedling length (r= 0.903), suggesting that more vigorous seedlings tend to be longer. The correlation between uniformity and length is moderate (r= 0.447), indicating a weaker association between these variables. Interaction between genotypes and treatments The interactions among hydropriming treatments within each genotype are illustrated in figure 5. The primed dry treatment had the least pronounced effect on vigor and uniformity, while primed stored allowed a partial recovery of these parameters. The interaction between genotypes within each treatment is shown in figure 6. Roxa exhibited a better response to the primed stored treatment, while Vanda demonstrated greater sensitivity. Seedling length was more affected in Vanda under the primed stored condition, whereas Roxa showed a tendency toward increased growth. Table 2 ­ Mean values of vigor index, uniformity index, and seedling length for control, primed dry, and primed stored treatments Hydropriming Vigor index Uniformity index Seedling length (cm) Control 783 770 3.63 Primed dry 727 745 3.29 Primed stored 762 765 3.37 Fig. 4 ­ Distribution, correlation, and dispersion analysis of the vigor index, uniformity index, and seedling length (cm). Fig. 5 ­ Interactions plot of hydropriming treatments within the Roxa and Vanda genotypes for the vigor index, uniformity index (both ranging from 0 to 1000), and seedling length (cm). Assumptions of MANOVA The suitability of the data for MANOVA was verified using the Henze­Zirkler and Anderson­ Darling tests (Tables 3 and 4). The Henze­Zirkler test indicated that the data did not present multivariate normality due to the length variable. Therefore, this variable was removed for the MANOVA analysis (Table 5). After its removal, multivariate normality was achieved (Table 6). The Box´s M test for equality of covariance matrices confirmed that the data met Fig. 6 ­ Interaction plot between Roxa and Vanda genotypes within each hydropriming treatment for the vigor index, uniformity index (both ranging from 0 to 1000), and seedling length (cm). Test Statistic p­value NMV* Henze­Zirkler 1.84 0.00000005 No Table 3 ­ Henze­Zirkler test for assessing the multivariate nor­ mality of the full dataset *Not Meeting Validity criteria (normality violated). Bewley, 2000; Marcos­Filho, 2015). The multivariate analysis showed that the Roxa genotype exhibited greater variability in vigor and seedling length, whereas Vanda demonstrated greater uniformity in growth. These results are consistent with studies indicating that different genotypes may exhibit significant variations in their physiological responses (Hampton and Tekrony, 1995; Elias et al., 2012; Rahman and Cho, 2016; Cheng et al., 2023). The significant interaction between hydropriming treatments and genotypes supports the hypothesis that the priming response may be cultivar­specific. The primed dry treatment had a weaker effect on vigor and uniformity, as reported in previous studies, which suggest that osmotic stress generated during the process may compromise seed physiological potential (Raj and Raj, 2019; Lewandowska et al., 2020; Pirasteh­Anosheh and Hashemi, 2020; Rhaman et al., 2020 a, 2020 b), particularly during the period immediately following treatment. Conversely, the primed stored treatment exhibited partial recovery of vigor and uniformity parameters, in agreement with studies highlighting the ability of seeds to 170 Adv. Hort. Sci., 2025 39(3): 165­173 the required assumptions for MANOVA (Table 7). A two­factor MANOVA was performed to evaluate the interaction effect between genotype and hydropriming treatment. The Roy’s largest root test indicated no significant differences between genotypes. However, the main effect of hydropriming treatment, as well as the interaction between genotype and treatment, were highly significant (Table 8). These results suggest that hydropriming treatments significantly influenced the analyzed variables, regardless of genotype. 4. Discussion and Conclusions The findings of this research can be understood in light of the existing literature on seed physiological potential and vigor. Seed vigor is one of the main factors influencing success seedling establishment in the field and early plant development (Black and Table 4 ­ Anderson­Darling test assessing the univariate normal­ ity of the vigor index, uniformity index, and seedling length variables Variable Statistic p­value Normality Vigor 0.6619 0.0830 Yes Uniformity 0.4255 0.3134 Yes Seedling length 15.860 0.0004 No *Not Meeting Validity criteria (normality violated). Table 5 ­ Henze­Zirkler test after removal of the seedling length variable, indicating multivariate normality of the remaining variables *Not meeting validity criteria (normality violated). Test Statistic p­value NMV* Henze­Zirkler 0.9217 0.1112 Yes Table 6 ­ Anderson­Darling test after removal of the seedling length variable, confirming the normality of the vigor index and uniformity index variables *Not Meeting Validity criteria (normality violated). Variable Statistic p­value Normality Vigor 0.6619 0.0830 Yes Uniformity 0.4255 0.3134 Yes Table 8 ­ MANOVA results for genotype and hydropriming treatment factors, considering the vigor index and uniformity index variables Table 7 ­ Box´s M test for the equality of covariance matrices among the analyzed groups Statistic p­value 2.05 0.56 Source DF Roy’s Statistic F Approximation Num. DF Den. DF p­value Genotype 1 0.002587 0.3014 2 233 0.74 Hydropriming 2 0.0726 85.044 2 234 0.0002721 *** Interaction 2 0.2100 245.726 2 234 0.00002 *** Residuals 234 Trujillo et al. ‐ Multivariate and digital analysis of lettuce seed vigor 171 stabilize after hydropriming when stored under appropriate conditions (Farooq et al., 2006, 2010; Huang et al., 2015; Souza et al., 2016; Farooq et al., 2021). Furthermore, the current results support research indicating that the effectiveness of hydropriming may vary depending on genotype and environmental conditions (Muhie et al., 2024). Recent studies emphasize the importance of evaluating each cultivar separately to determine the most suitable seed treatment method (Cheng et al., 2023; Qiu et al., 2023). The positive correlation between vigor and seedling length reinforces the relevance of these variables in seed quality assessment. The literature suggests that more vigorous seedlings tend to develop stronger root systems and exhibit higher field emergence rates (Kikuti and Marcos­Filho, 2012; Kikuti and Marcos­Filho, 2013; Alvarenga and Marcos­Filho, 2014; Marcos­Filho, 2015; Rego et al., 2023). Prior studies indicate that seed vigor is closely associated with early seedling growth and crop establishment (Marcos­Filho, 2015). Image analysis has proven to be a promising tool for evaluating seed vigor. Technologies such as the Seed Vigor Imaging System (SVIS®) have demonstrated a high degree of precision in classifying lettuce seed lots and those other crops (Gomes­Junior et al., 2009; Rodrigues et al., 2020). The use of computer vision and machine learning in seed vigor assessment is increasingly being explored, enabling fast and objective analyses (De Medeiros et al., 2020; Wang et al., 2021; Liu et al., 2023; Pang et al., 2023). The statistical methodology adopted in this research was essential to ensure the robustness of the analyses and the reliability of the results. Initially, the descriptive analysis enabled the identification of trends and patterns in the data, facil itating interpretation. To assess relationships among variables, Pearson´s correlation was applied, revealing a strong association between the vigor index and seedling length. The main statistical method used was Multivariate Analysis of Variance (MANOVA), a widely accepted methodology approach studies involving correlated dependent variables (Oliveira et al., 2013; Din and Hayat, 2021; Baumeister et al., 2024). MANOVA is particularly suitable when response variables are correlated, allowing for the simultaneous evaluation of the effects of experimental factors (Johnson and Wichern, 2002). To ensure the method’s applicability, the Henze­ Zirkler test was used to assess multivariate normality, and Box’s M test was applied to verify the homogeneity of covariance matrices­a key assumption for valid MANOVA results. The significance of main effects and interactions was assessed using Roy’s largest root, which is recommended when effects have a strong impact on data variability (Kose et al., 2018). The MANOVA results were complemented by univariate analyses, allowing for a more detailed interpretation of the individual factor effects, as suggested by Scholz and Stephens (1987). This combined approach improves precision in identifying significant effects and interactions, thereby enhancing the understanding of genotype responses to hydropriming. However, such statistical procedures also have limitations. In this study, seedling length had to be excluded from the final MANOVA due to the violation of normality assumptions, which restricted the scope of multivariate interpretation. In conclusion, the Roxa genotype performed better under the primed stored treatment than Vanda. Seedling length was influenced by hydropriming, with primed stored proving unsuitable for Vanda. Therefore, the statistical approach adopted in this study enabled a comprehensive and detailed analysis, enhancing our understanding of the effects of hydropriming on the evaluated genotypes. These findings are crucial for understanding genotypes­ treatments interactions and may contributing to the optimization of hydropriming strategies for lettuce seeds. Future studies should consider incorporating a broader range of cultivars, extended storage durations, and the integration of machine learning techniques to improve vigor prediction. Acknowledgements To the ‘Ministerio de Ciencia, Tecnología e Innovación (MINCIENCIAS) / Colfuturo’, Colombia, for supporting this research through the ‘Doctorado Exterior’ ­ 885 scholarship since 2021. To the ‘Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)’, Brazil, for the ‘PROEX’ scholarship granted throughout the doctoral program. References ALVARENGA R.O., MARCOS­FILHO J., 2014 ‐ Vigor evaluation of stored cotton seeds, including the Seed Vigor Imaging System (SVIS®). ­ J. Seed Sci., 36(2): 222­ 230. ALVARENGA R.O., MARCOS­FILHO J., TIMÓTEO T.S., 2013 ­ Assessment of the physiological potential of super sweet corn seeds. ­ J. Seed Sci., 35(3): 340­346. ANDERSON T.W., 2003 ­ An introduction to multivariate statistical analysis. ­ Wiley & Sons, Inc., 3rd edition. Hoboken, NJ, USA, pp. 752. BAUMEISTER M., DITZHAUS M., PAULY M., 2024 ­ Quantile‐based MANOVA: A new tool for inferring multivariate data in factorial designs. ­ J. Multivar. Anal., 199: 105246. BLACK M., BEWLEY J.D., 2000 ­ Seed technology and its biological basis. 1st ed., Vol. 1. ­ CRC Press, Boca Raton, FL, USA, pp. 410. CALDEIRA C.M., MOREIRA DE CARVALHO M.L., OLIVEIRA J.A., VILAS BOAS COELHO S., YUMI KATAOKA V., 2014 ­ Vigor de sementes de girassol pela análise computadorizada de plântulas [Sunflower seed vigor by computerized seedling analysis]. ­ Científica, 42(4): 346­ 353. CASEIRO R.F., 2003 ­ Métodos para o condicionamento fisiológico de sementes de cebola e influência da secagem e armazenamento [Methods for the physiological conditioning of onion seeds and influence of drying and storage]. ­ Escola Superior de Agricultura Luiz de Queiroz, Univ. São Paulo, Piracicaba, SP, Brasil. CHENG T., CHEN G., WANG Z., HU R., SHE B., PAN Z., ZHOU X.­G., ZHANG G., ZHANG D., 2023 ­ Hyperspectral and imagery integrated analysis for vegetable seed vigor detection. ‐ Infrared Phys. Technol., 131: 104605. DE MEDEIROS A.D., CAPOBIANGO N.P., DA SILVA J.M., DA SILVA L.J., DA SILVA C.B., DOS SANTOS DIAS D.C.F., 2020 ­ Interactive machine learning for soybean seed and seedling quality classification. ­ Sci. Rep., 10: 11267. DIN I., HAYAT Y., 2021 ­ ANOVA or MANOVA for correlated traits in agricultural experiments. ­ Sarhad J. Agric., 37(4): 1250­1259. ELIAS S.G., COPELAND L.O., MCDONALD M.B., BAALBAKI R.Z., 2012 ­ Seed testing: principles and practices. ­ Michigan State Univ. Press, 1st ed., Vol. 1. East Lansing, MI, USA, pp. 368. FAROOQ M., BASRA S.M., WAHID A., AHMAD N., 2010 ­ Changes in nutrient‐homeostasis and reserves metabolism during rice seed priming: consequences for seedling emergence and growth. ­ Agric. Sci. China, 9(2): 191­198. FAROOQ M., BASRA S.M.A., AFZAL I., KHALIQ A., 2006 ­ Optimization of hydropriming techniques for rice seed Adv. Hort. Sci., 2025 39(3): 165­173 172 invigoration. ‐ Seed Sci. Technol., 34(2): 507­512. FAROOQ M., ROMDHANE L., REHMAN A., AL­ALAWI A.K.M., AL­BUSAIDI W.M., ASAD S.A., LEE D.­J., 2021 ­ Integration of seed priming and biochar application improves drought tolerance in cowpea. ­ J. Plant Growth Regul., 40(5): 1972­1980. GOMES­JUNIOR F.G., CHAMMA H.M.C.P., CICERO S.M., 2014 ­ Automated image analysis of seedlings for vigor evaluation of common bean seeds. ­ Acta Sci. Agron., 36(2): 195­202. GOMES­JUNIOR F.G., MONDO V.H.V., CICERO S.M., MCDONALD M.B., BENNETT M.A., 2009 ‐ Evaluation of priming effects on sweet corn seeds by SVIS®. ­ Seed Technol., 31(1): 95­100. HAMPTON J.G., TEKRONY D.M., 1995 ­ Handbook of vigour test methods. 3rd ed. ­ International Seed Testing Association, Zurich, Switzerland, pp. 117. HAND D.J., TAYLOR C.C., 1987 ­ Multivariate analysis of variance and repeated measures: a practical approach for behavioural scientists. Vol. 5. ­ CSR Press, New York, NY, USA. HENZE N., ZIRKLER B., 1990 ­ A class of invariant consistent tests for multivariate normality. ­ Commun. Stat. Theory Methods, 19(10): 3595­3617. HOFFMASTER A.F., FUJIMURA K., MCDONALD M.B., BENNETT M.A., 2003 ­ An automated system for vigor testing three‐day‐old soybean seedlings. ­ Seed Sci. Technol., 31(3): 701­713. HUANG M., WANG Q.G., ZHU Q.B., QIN J.W., HUANG G., 2015 ­ Review of seed quality and safety tests using optical sensing technologies. ­ Seed Sci. Technol., 43(3): 337­366. HUBERTY C.J., OLEJNIK S., 2006 ­ Applied MANOVA and discriminant analysis. 2nd ed. ­ Wiley, Hoboken, NJ, USA. JOHNSON R.A., WICHERN D.W., 2002 ­ Applied multivariate statistical analysis (6th ed.). ­ Prentice Hall, Upper Saddle River. KIKUTI A.L.P., MARCOS­FILHO J., 2012 ­ Testes de vigor em sementes de alface [Vigor tests in lettuce seeds]. ­ Horticult. Bras., 30(1): 44­50. KIKUTI A.L.P., MARCOS­FILHO J., 2013 ­ Análise de imagens de plântulas e testes tradicionais para avaliação do vigor de sementes de quiabo [Seedling image analysis and traditional tests for assessing okra seed vigor]. ­ J. Seed Sci., 5(4): 443­448. KOSE A., ONDER O., BILIR O., KOSAR F., 2018 ­ Application of multivariate statistical analysis for breeding strategies of spring safflower (Carthamus tinctorius L.). ­ Turk. J. Field Crops, 23(1): 12­19. KRZANOWSKI W.J., 2000 ­ Principles of multivariate analysis: A User’s Perspective. ­ Ford Statistical Science Series, Oxford University Press, Oxford, UK, pp. 608. KRZANOWSKI W.J., MARRIOTT F.H.C., 1994 ­ Multivariate analysis: Distributions, ordination and inference. Vol. Trujillo et al. ‐ Multivariate and digital analysis of lettuce seed vigor 173 1). Edward Arnold, London, UK, pp. 280. LEWANDOWSKA S., ŁOZIŃSKI M., MARCZEWSKI K., KOZAK M., SCHMIDTKE K., 2020 ­ Influence of priming on germination, development, and yield of soybean varieties. ­ Open Agric., 5(1): 930­935. LIU F., YANG R., CHEN R., LAMINE GUINDO M., HE Y., ZHOU J., LU X., CHEN M., YANG Y., KONG W., 2023 ­ Digital techniques and trends for seed phenotyping using optical sensors. ­ J. Adv. Res., 63: 1­16. MARCHI J.L., CICERO S.M., GOMES­JUNIOR F.G., 2011 ­ Using computerized analysis of seedlings to evaluate the physiological potential of peanut seeds treated with fungicide and insecticide. ­ Rev. Bras. Sementes, 33(4): 652­662. MARCOS­FILHO J., 1999 ­ Testes de vigor: importância e utilização [Vigor tests: importance and use], pp. 1­21. ­ In: KRZANOWSKI F.C., R.D. VIEIRA, and FRANÇA NETO J.B. (eds.) Vigor de sementes: conceitos e testes (Vol. 1,). ABRATES, Londrina, Brazil, pp. 601. MARCOS­FILHO J., 2015 ­ Fisiologia de sementes de plantas cultivadas [Physiology of seeds of cultivated plants]. ABRATES, 2nd edition, Londrina, Brazil, pp. 660. MARCOS­FILHO J., BENNETT M.A., MCDONALD M.B., EVANS A.F., GRASSBAUGH E.M., 2006 ­ Assessment of melon seed vigour by an automated computer imaging system compared to traditional procedures. ­ Seed Sci. Technol., 34(2): 485­497. MARCOS­FILHO, J., 2015 ­ Seed vigor testing: an overview of the past, present and future perspective. ­ Sci. Agric., 72(4): 363­374. MUHIE S.H., AKELE F., YESHIWAS T., 2024 ­ Phenological and yield response of primed carrot (Daucus carota L.) seeds under deficit irrigation. ­ Adv. Hort. Sci., 38(2): 119­127. NICACIO J.E.M., PERUSSOLO M.A., LIMA A.C.S. S., 2013 ­ Análise de variância multivariada‐manova na seleção de produtores de laranja Citrus sinensis (L.) Osbeck [Multivariate analysis of variance­MANOVA in the selection of sweet orange producers]. ­ Rev. Estud. Soc., 15(30): 189­202. OLIVEIRA I.R.C., REZENDE M.T., DIAS C.T.S., GOMES D.S., BOTREL É.P., GOMES L.A. A., 2013 ­ Evaluation of crisphead lettuce cultivars in different cover types by MANOVA and discriminant analysis. ­ Horticult. Bras., 31(3): 439­444. PANG T., CHEN C., FU R., WANG X., YU H., 2023 ­ An end‐ to‐end seed vigor prediction model for imbalanced samples using hyperspectral image. ­ Front. Plant Sci., 14: 1­16. PIRASTEH­ANOSHEH H., HASHEMI S.­E., 2020 ­ Priming, a promising practical approach to improve seed germination and plant growth in saline conditions. ­ Asian J. Agric. Food Sci., 8(1): 1­12. QIU C., DING F., HE X., WANG M., 2023 ­ Apply physical system model and computer algorithm to identify Osmanthus fragrans seed vigor based on hyperspectral imaging and convolutional neural network. ­ Inf. Technol. Control, 52(4): 887­897. R CORE TEAM, 2024 ­ R: A language and environment for statistical computing (4.3.3). ­ R Foundation for Statistical Computing. RAHMAN A., CHO B.­K., 2016 ­ Assessment of seed quality using non‐destructive measurement techniques: a review. ­ Seed Sci. Res., 26(4): 285­305. RAJ A.B., RAJ S.K., 2019 ­ Seed priming: an approach towards agricultural sustainability. ­ J. Appl. Nat. Sci., 11(1): 227­234. REGO C.H.Q., BRITO D.L., TORRES S.B., MORAIS E.R.C., PEREIRA M.D., DUTRA A.S., BACHETTA G., ALVES C. Z., 2023 ­ Primary root emission as a vigor test in soybean seeds. ­ Rev. Ciênc. Agron., 54: , e20238714. RHAMAN M.S., IMRAN S., RAUF F., KHATUN M., BASKIN C.C., MURATA Y., HASANUZZAMAN M., 2020 a ­ Seed priming with phytohormones: an effective approach for the mitigation of abiotic stress. ­ Plants, 10(1): 37. RHAMAN M.S., RAUF F., TANIA S.S., KHATUN M., 2020 b ­ Seed priming methods: application in field crops and future perspectives. ­ Asian J. Res. Crop Sci., 5(2): 8­19. ROCHA C.R.M., DA SILVA V.N., CICERO S.M., 2015 ­ Sunflower seed vigor evaluation by seedling image analyze. ­ Cienc. Rural, 45(6): 970­976. RODRIGUES M., GOMES­JUNIOR F.G., MARCOS­FILHO J., 2020 ­ Vigor‐S: system for automated analysis of soybean seed vigor. ­ J. Seed Sci., 42: e202042039. SAKO Y., MCDONALD M. B., FUJIMURA K., EVANS A.F., BENNETT M., 2001 ­ A system for automated seed vigour assessment. ­ Seed Sci. Technol., 29: 625–636. SCHOLZ F.W., STEPHENS M.A., 1987 ­ K‐sample Anderson– Darling tests. ­ J. Am. Stat. Assoc., 82(399): 918­924. SOUZA P.P., MOTOIKE S.Y., CARVALHO M., KUKI K.N., BORGES E.E.L.E, SILVA A. M., 2016 ­ Storage on the vigor and viability of macauba seeds from two provenances of Minas Gerais State. ­ Cienc. Rural, 46(11): 1932­1937. WANG C., LIU B., LIU L., ZHU Y., HOU J., LIU P., LI X., 2021 ­ A review of deep learning used in the hyperspectral image analysis for agriculture. ­ Artif. Intell. Rev., 54(7): 5205­5253. WATERS­JUNIOR L., BLANCHETTE B., 1983 ­ Prediction of sweet corn field emergence by conductivity and cold tests. ­ J. Am. Soc. Hort. Sci., 108(5): 778­781. WILKS S.S., 1935 ­ On the independence of k sets of normally distributed statistical variables . ­ Econométrica, 3(3): 309. XIA Y., XU Y., LI J., ZHANG C., FAN S., 2019 ­ Recent advances in emerging techniques for non‐destructive detection of seed viability: A review. ­ Artif. Intell. Agric., 1: 35­47.