Computer-aided digital image analysis of seedling size and growth rate for assessing seed vigour in Impatiens

Oakley, K., Kester, S.T. and Geneve, R.L. (2004), Seed Sci. & Technol., 32, 907-915 Computer-aided digital image analysis of seedling size and growth rate for assessing seed vigour in Impatiens K. OAKLEY, S.T. KESTER AND R.L. GENEVE* Department of Horticulture, University of Kentucky, Lexington, KY 40546, University of Kentucky, USA (E-mail: Rgeneve@uky.edu.) (Accepted October 2003) Summary Vigour was measured using computer-aided analysis of digital images in six seed lots of Impatiens that differed in vigour but retained greater than 86% standard germination. Seed lots that differed in initial seed vigour were selected based on the commercially used Ball Vigour Index and vigour independently assessed using saturated salts accelerated aging tests. Digital images were captured from seeds germinated in Petri dishes placed on a flat-bed scanner. Seedling growth was measured daily for four days following initial radicle protrusion using commercially available root length calculating software. Seedling size and growth rate generally ranked seed lots from high to low vigour in a similar way compared to the Ball Vigour Index and saturated salts accelerated aging tests. The exception was that seed lot #2 was identified as the highest vigour seed lot using seedling growth parameters rather than seed lot #1. However, high and low vigour seed lots were identified as well with seedling growth measurements compared to the other two vigour tests used to evaluate Impatiens seed lots. Additionally, two vigour indices were calculated for seedling length using standard deviation as a measure of population uniformity. Both indices showed less ability to statistically separate seed lots compared to growth measurements alone. The current study showed that computer-aided analysis of digital images could be used successfully to rank seed lot vigour in Impatiens based on seedling length. Introduction Vigour is the ability of a seed lot to establish normal (or usable) seedlings under diverse production environments (Hampton and TeKrony, 1995). It is routinely evaluated in commercial seed lots as part of quality assurance programs. Plug production for bedding plants has become an automated and technically sophisticated production system that relies on high quality seeds for consistent results (Styer and Koranski, 1997). Flower seeds constitute a significant investment for plug and bedding plant growers and even though germination is initiated under relatively controlled environments, successful stand establishment can vary depending on a seed lot s initial vigour (Cantliffe, 1998). Vigour tests are essential for seed production companies to evaluate and develop production and post-harvest techniques, to make inventory management and sales decisions and to justify premium prices. Vigour tests are also essential for commercial growers who need to have the means to independently assess seed vigour, to justify buying premium-priced seed and to have confidence in the performance of their crops (Karlovich, 1998). Therefore, it is * Author for correspondence 907

K. OAKLEY, S.T. KESTER AND R.L. GENEVE important that standard seed vigour tests become available for flower seed producers and consumers. The International Seed Testing Association (Perry, 1981) and Association of Official Seed Analysts (AOSA, 1983) have published guidelines for vigour testing in seeds, which have been updated by Hampton and TeKrony (1995). Vigour tests are grouped into three categories that include: 1) single tests based on germination behavior; 2) physiological or biochemical indices of vigour; and 3) multiple testing procedures. Single seed vigour tests include measures of seedling growth or normal germination percentage after a stress imposition. Seedling growth tests include measures of germination rate (time to radicle protrusion), early seedling growth (size or rate) and sorting seedlings into strong or weak growing categories. Vigour tests that evaluate normal germination after stress include accelerated aging, controlled deterioration, cool and cold tests. Biochemical tests include various measures of metabolic activity in seeds. These include aspects related to respiration (i.e. ATP synthesis, tetrazolium staining), membrane repair (i.e. electrolyte leakage) and volatile production (i.e. ethylene). Of these tests, only electrolyte leakage measured by conductivity tests has been used extensively to evaluate commercial seed lot vigour (McDonald, 1998). Multiple testing addresses the premise that seed vigour is a complex response to variable environments and that all seeds in a seed lot may not respond the same to those environments. Therefore, multiple testing uses several different vigour tests in assessing the overall vigour of a seed lot. The primary vigour tests for large-seeded agronomic crops are standard germination following imposition of stress, including accelerated aging and the cold test (Hampton and TeKrony, 1995). However, for smaller seeded crops such as vegetable and flowers, the only stress tests that have been consistently applied successfully include controlled deterioration (Powell and Matthews, 1981) and a modified accelerated aging test termed saturated salts accelerated aging (Jianhua and McDonald, 1996). Use of these tests has demonstrated how important managing water uptake during the stress test is to conducting a successful test. Alternatives to stress tests for seed vigour assessment in small-seeded crops are seedling growth measurements after germination. These are attractive vigour tests because they rely on the simple premise that seedling growth following radicle protrusion is correlated to seed vigour and that this can be universally applied to all small-seeded crops. Typical systems germinate seeds on a slant board to generate straight radicles that are measured by hand (Smith et al., 1973). Gray and Steckel (1983) showed that seedling length measured by the slope test predicted field emergence in carrot as well as a controlled deterioration test. Because seed size is small for most flower crops, early seedling growth can be difficult and time consuming to measure and results can easily vary between analysts (Geneve and Kester, 2001). Use of computer-aided image analysis of seedling size overcomes many of these limitations (McDonald et al., 2001). Image analysis provides rapid measurement of an object s physical characteristics and allows quantitative, objective observation (Kranzler, 1985). Several commercial systems use some form of computer-aided analysis of digital images to evaluate seedling growth as a measure of seed vigour. These include the 908

Paradigm Seed lot Vigour Assessment System (McNertney, 1999) and the Ball Vigour Index (Conrad, 1999) that use a CCD video camera to capture images of seedlings. Subsequently, Geneve and Kester (2001) and Sako et al. (2001) developed similar systems that capture images using a flat-bed scanner. Software has been developed to evaluate seedling growth and in most cases a vigour index has been computed based on seedling growth, growth uniformity and germination percentage (Conrad, 1999; McNertney, 1999; Sako et al., 2001). However, there is no published evidence that seedling size or growth rate measured by computer-aided analysis correlates with other standard measures of vigour for small-seeded crops, nor is there evidence that a vigour index value provides more information about seed lot vigour compared to simple growth analysis. Therefore, the objective of the current study was to use computer-aided image analysis to evaluate several measures of seedling growth in seed lots of Impatiens with different vigour, but high germination percentages. Specifically, seed lots from a single Impatiens cultivar were selected based on their vigour as determined by Ball Seed Company (East Chicago, IL), using their Ball Vigour Index and verified using the saturated salts accelerated aging vigour test (Jianhua and McDonald, 1996). Measures to be evaluated included seedling growth (length and area), growth rate and a vigour index, calculated as the sum of seedling growth plus uniformity. Methods and materials Six seed lots of Impatiens wallerana Hooker F. Super Elfin Pink were obtained from Ball Seed, Inc. (East Chicago, IL). Vigour of each lot had been measured by Ball Seed using the Ball Vigour Index based on cotyledon area of seedlings produced in plug trays under controlled environmental conditions (Conrad, 1999). Seed size on a weight and area basis was not different between seed lots (data not shown). Upon receipt, seeds were stored in airtight foil packages at 5 C. Standard germination was conducted for 25 seeds placed in four replicate 8.5-cm diameter plastic petri dishes containing one piece of blue blotter (Anchor Paper Co., St. Paul, Minn.) with 5.25 ml of water sealed with Parafilm (American National Can, Menasha, Wis.) and placed in a germination chamber held at a constant 23 C with 40 µmol s -1 m -2 from cool-white fluorescent lamps. Normal seedlings were counted 18 days after sowing. Each germination test was repeated. Saturated salt accelerated aging vigour tests were conducted using one-hundred seeds of each seed lot suspended over a saturated solution of NaCl in accelerated aging boxes and placed in a water jacketed accelerated aging chamber and held at 40 C for 48 hr (Jianhua and McDonald, 1996). Germination percentages were evaluated using standard germination conditions immediately after saturated salt accelerated aging. Seed moisture content on a fresh weight basis was measured before and after accelerated aging. Each vigour test was repeated. Seedling growth was measured in four replicates containing nine seeds per seed lot sown in 8.5-cm diameter plastic Petri dishes containing one piece of clear, uncoated cellulose film (Celorey-PUT, Cydsa Monterrey, Mexico). The transparent cellulose 909

K. OAKLEY, S.T. KESTER AND R.L. GENEVE film was cut to fit the Petri dish, soaked in distilled water for 30 min to remove surface contamination and autoclaved for 15 min to eliminate microbes from the film. Water (1.5 ml) was added to the dish prior to placing seeds on the film. Petri dishes were sealed with Parafilm and placed in a single germination chamber held at a constant 23 C at 40 µmol s -1 m -2 from cool-white fluorescent lamps. Seedling growth was evaluated using computer-aided analysis of digital images (Geneve and Kester, 2001). Petri dishes were scanned and digital images acquired using a flat-bed scanner (Hewlett Packard Scanjet 4c/5, Palo Alto, Calif.) that included both base and top lighting. Each scan created a 300 dpi (118 dots per cm) resolution, black and white image. Petri dishes were scanned with lids on once every 24 h from 4 to 7 days after imbibition. Seedling length (cm), area (mm 2 ) and growth rate (calculated from the slope of the linear regression equation) were measured over a four day period using MacRhizo (Regent Instruments, Inc., Quebec, Canada) with a threshold setting of 200. Two vigour indices were calculated using standard deviation as an indicator of uniformity. Vigour index #1 was calculated for each replicate (Petri dish) by summing seedling length after 7-d with 1 minus the standard deviation (Sako et al., 2001). Vigour index #2 was similar to the Ball Vigour Index (Conrad, 1999) and calculated by dividing the mean 7-d seedling length for a Petri dish by the standard deviation followed by multiplying this value by the germination percentage. Seeds were considered germinated if the radicle exceeded 3 mm. A total of 36 seeds (four Petri dishes with nine seeds) were evaluated for each seed lot and each experiment was repeated. Results All six Impatiens seed lots had statistically similar high standard germination percentages greater than 96% (table 1). While standard germination did not reveal differences between seed lots, the Ball Vigour Index indicated that two seed lots (1, 2) had a vigour index greater than 600, two seed lots (3, 4) had a vigour index between 500 and 600 and two seed lots (5, 6) had a vigour index below 500. The saturated salt accelerated aging test statistically distinguished the same two seed lots (1, 2) as high vigour, one seed lot clearly being a medium vigour seed lot (4) and two low vigour seed lots (5, 6). Seed lot 3 was not significantly different between the medium and low vigour seed lots using the saturated salt accelerated aging test. Computer-aided measurement of seedling length statistically identified three levels of seedling vigour (table 1), one high performing seed lot (2), two seed lots with moderate vigour (1, 3) and three low vigour seed lots (4, 5, 6). Seedling area measurements were highly correlated to seedling length (r 2 = 0.82) and therefore ranked seed lots by vigour in the same order as seedling length (table 1). However, seedling area statistically grouped seed lot 3 from a medium to high vigour category and seed lot 4 from low to medium. Seedling growth rate for length or area was linear over the first seven days for each seed lot regardless of vigour (figure 1). Seedling length growth rate identified one high vigour (2), two medium vigour (1, 3) and three low vigour (4, 5, 6) seed lots, while area growth rate identified two high (2, 3), two medium (1, 4) and two low vigour (5, 6) 910 F O R M A T T E D P R O O F

Table 1. Standard germination and seed vigour determined by several methods for impatiens seeds varying in initial seed quality. Seed lot Standard Germination (%) Ball Vigour Index Saturated Salts Accelerated aging (%) Seedling size after 7 days Length (cm) Area (mm 2 ) 1 96 a 651 81 a z 1.05 b 0.74 b 2 97 a 642 84 a 1.44 a 1.04 a 3 97 a 561 69 bc 1.15 b 0.94 a 4 96 a 505 75 b 0.70 c 0.71 b 5 98 a 485 54 c 0.70 c 0.59 c 6 96 a 440 48 c 0.61 c 0.48 c z Means with the same letter within a column were not different by Tukey s test (5%). 1.5 Seed lot number Growth rate Seedling area (mm 2 ) Seedling length (cm) 1.0 0.5 0.0 1.0 0.5 1 2 3 4 5 6 0.36 a 0.26 b 0.25 b 0.13 c 0.13 c 0.12 c Growth rate 0.23 a 0.22 a 0.14 b 0.13 b 0.09 c 0.08 c 0.0 6 7 8 9 Days after imbibtion Figure 1. Seedling length (cm ± SE) and growth rate (cm per day ) and seedling area (mm 2 ± SE) and growth rate (mm 2 per day) for six impatiens seed lots of varying seed quality. The r 2 for all seed lots was 0.92. 911

K. OAKLEY, S.T. KESTER AND R.L. GENEVE seed lots. Seedling growth rate was similar to vigour levels identified by the saturated salts accelerated aging test and the Ball Vigor Index. Seedling growth rates, however, consistently found seed lot 2 to be the highest vigour level and seed lot 3 ranked higher than in saturated salts accelerated aging and the Ball Vigor Index tests. Using vigour indices resulted in a reduction in the ability to statistically separate seed lots compared to using length alone (table 2). Vigour index #1 statistically ranked seed lots the same as seedling length alone but was unable to statistically separate seed lots 2 and 3. Vigour index #2 was only able to separate seed lot 3 from seed lot 6. Errors in computer-aided seedling lengths were evident when the hypocotyl hook was not fully open (figure 2). Growing seedlings on a slant board could reduce errors, but the overall results did not change significantly (data not shown). Table 2. Seed vigour using seedling length alone or two vigour indices. Seed lot Seedling length Vigour index #1 z Vigour index #2 y 1 1.05 b 1.61 b w 248.5ab 2 1.44 a 1.96 a 311.5ab 3 1.15 b 1.78 ab 391.1a 4 0.70 c 1.47 c 325.4ab 5 0.70 c 1.47 c 307.7ab 6 0.61 c 1.36 c 219.9b z Vigour index #1 is the sum of the length after 7-d plus 1- standard deviation. y Vigour index #2 is the sum of the length after 7-d divided the standard deviation times the germination percentage. w Means with the same letter within a column were not different by Tukey s test (5% level). A B C Hypocotyl loop Hypocotyl shortcut Accurate trace Figure 2. Examples of traced images of impatiens. The computer creates a single pixel width skeleton to estimate the length of the seedling. Each panel includes: the original image; the skeleton trace overlayed on the image; and the skeleton trace alone. (A) the hypocotyl loop trace overestimates seedling length. (B) the hypocotyl shortcut underestimates the seedling length. (C) is an accurate trace of the seedling including the adventitious roots emerging from the hypocotyl. 912

Discussion The current study supports the use of early seedling growth and early seedling growth rates as appropriate measures of seed vigour. Radicle length has been used to test vigour in a number of small-seeded crops including carrot, lettuce (McCormac et al., 1990; Smith et al., 1973), cauliflower, onion and leek (Finch-Savage, 1986). Seed lots in the current study were consistently ranked with both a vigour stress test (SSAA) previously used to evaluate vigour in Impatiens (Jianhua and McDonald, 1996) and the commercial Ball Vigour Index, which measures seedling emergence to indicate vigour (Conrad, 1999). Seedling growth only differed from these two tests by identifying seed lot 2 as the highest vigour lot rather than seed lot 1. However, high and low vigour seed lots were identified equally as well with seedling growth measurements compared to the other two vigour tests used to evaluate Impatiens seed lots. Seedling growth rate is thought to be a sensitive measure of seed vigour (Woodstock, 1969), but is difficult to incorporate in routine vigour testing because it has been too labor intensive to periodically evaluate seedling growth over time. Computer-aided analysis of digital images reduces the labor input required to evaluate seedling growth rate and increased the accuracy of these measurements (Geneve and Kester, 2001; Sako et al., 2001). Early growth rate in Impatiens is linear over time and, therefore, did not differ significantly for evaluating vigour compared to a single length measurement taken after seven days (table 1 and figure 1). Since a single measurement requires less labor compared to repeated measurements, the single assessment of seedling length was adequate for vigour testing in Impatiens. Intuitively, uniformity would seem to be an essential component of seed vigour. Rapid and uniform stand establishment is one objective for evaluating seed lots for vigour. Orchard (1977) incorporated standard deviation of emergence times as a uniformity parameter in a vigour index. Similarly, Conrad (1999) and Sako et al. (2001) used standard deviation to calculate vigour indices for small-seeded crops. Adapting these indices to the data collected for seedling length in Impatiens resulted in less statistical separation between seed lots (table 2). Separation of seed lots was further reduced if growth or uniformity were assigned weighted values (data not shown) as suggested by Sako et al. (2001). These data do not necessarily discount uniformity as an important aspect of seed vigour, but does indicate that the inclusion of standard deviation with length measurements does not improve the ability to separate seed lots for vigor based on seedling length. The poorer resolution using standard deviation was probably due to the proportionally small standard deviations associated with the shorter lengths observed in the slower growing seedlings from lower vigour seed lots. Also, the high germination percentages for seed lots used for the current study negated its use in vigour indices calculations. Errors were encountered for evaluating seedling length using computer-aided analysis (figure 2). Errors occurred most often when the hypocotyl hook was tight and the cotyledons close to the hypocotyl. This resulted in either a hypocotyl loop that overestimated seedling size or a hypocotyl short cut that underestimated seedling size. The Rhizo software used to automatically measure seedling growth in this study was 913

K. OAKLEY, S.T. KESTER AND R.L. GENEVE designed to measure root systems. It allows for manual correction of these errors, but this requires analyst input and negates some of the benefits in labor savings associated with computer-aided analysis. Hoffmaster (2002) observed similar calculation errors for computer-aided seedling length in soybean (Glycine) and designed custom software to identify and correct typical errors prior to calculating length. The percentage of miscalculation of seedling length observed in Impatiens seed lots was relatively small (< 5%) and would not significantly alter seedling vigour rankings. In a previous study using the same computer-aided analysis system to evaluate several small-seeded crop species, there were no statistical differences comparing hand measurements with computer analysis suggesting that both methods were subject to similar variation (Geneve and Kester, 2001). A slant board was not used in this study, but orienting Petri dishes to provide a geotropic stimulus tended to increase the distance between cotyledons and hypocotyl reducing the incidence of computational errors (data not shown). In conclusion, the current study showed that computer-aided analysis of digital images successfully ranked seed lot vigour in Impatiens seed lots based on seedling growth. There was no difference between measuring seedling growth after seven days or using growth rate. Incorporating a uniformity factor into a vigour index calculation reduced the ability to separate seed lots for vigour and does not appear to be appropriate for vigour testing in Impatiens. Acknowledgements Seeds for this study were generously supplied by Ball Seed, Inc. This research was partially funded through a grant from the Gloeckner Foundation. References Association of Official Seed Analysts (1983). Seed Vigour Testing Handbook, Contribution No. 32. Association of Official Seed Analysts, Lincoln, Nebr. Cantliffe, D.J. (1998). Seed germination for transplants. HortTechnology, 8, 499 503. Conrad, R. (1999). Method and apparatus for assessing the quality of a seed lot. U. S. Patent # 5,901,237. Finch-Savage, W.E. (1986). A study of the relationship between seedling characters and rate of germination within a seed lot. Annals of Applied Biology, 108, 441 444. Geneve, R.L. and Kester, S.T. (2001). Evaluation of seedling size following germination using computer-aided analysis of digital images from a flat-bed scanner. HortScience, 36, 1117 1120. Gray, D. and Steckel, J.R.A. (1983). A comparison of methods for evaluating seed quality in carrots (Daucus carota). Annals of Applied Biology, 103, 327 334. Hampton, J.G. and TeKrony, D.M. (1995). Vigour Testing Methods, 3 rd edition. International Seed Testing Association. Zurich. Hoffmaster, A.L. (2002). An automated system for seed vigour testing three-day-old soybean seedlings. M.S. Thesis, The Ohio State University, Columbus, OH, USA. Jianhua, Z. and McDonald, M.B. (1996). The saturated salt accelerated aging test for small-seeded crops. Seed Science and Technology, 25, 123 131. Karlovich, P.T. (1998). Flower seed testing and reporting needs of the professional grower. Seed Technology, 20, 131 135. 914

Kranzler, G.A. (1985). Applied digital image processing in agriculture. Agricultural Engineering, 66, 11 13. McCormac, A.C., Keefe, P.D. and Draper, S.R. (1990). Automated vigour testing of field vegetables using image analysis. Seed Science and Technology, 18, 103 112. McDonald, M.B. (1998). Seed quality assessment. Seed Science Research, 8, 265 275. McDonald, M.B., Evans, A.F. and Bennett, M.A. (2001). Using scanners to improve seed and seedling evaluations. Seed Science and Technology, 29, 683 689. McNertney, D.C. (1999). System and method for measuring seedlot vigour. U. S. Patent # 5,864,984. Orchard, T.J. (1977). Estimating the parameters of plant seedling emergence. Seed Science and Technology, 5, 61 69. Perry, D.A. (1981). Handbook of Vigour Test Methods, International Seed Testing Association, Zurich, Switzerland. Powell, A.A. and S. Matthews (1981). Evaluation of controlled deterioration, a new vigour test for small seeded vegetables. Seed Science and Technology, 9, 633 640. Sako, Y., McDonald, M.B., Fujimura, K., Evans, A.F. and Bennett, M.A. (2001). A system for automated seed vigour assessment. Seed Science and Technology, 29, 625 636. Smith, O.E., Welch, N.C. and Little, T.M. (1973). Studies on lettuce seed quality, I. Effect of seed size and weight on vigour. Journal of the American Society for Horticultural Science, 98, 529 533. Styer, R.C. and Koranski, D.S. (1997). Plug and Transplant Production, Ball Publishing, Batavia, IL. Woodstock, L.W. (1969). Seedling growth as a measure of seed vigour. Proceedings of the International Seed Testing Association, 34, 273 280. 915