The SSAA test was used as a seed vigor test. The SSAA value (%) represented resistance to hot and humid conditions by the seed. The controlled conditions not only maintained germination of rice seed but also could delay a reduction of seed vigor for 6 months (Fig. 2a). However, storage under ambient, conditions decreased seed vigor rapidly (Fig. 2b). Interestingly, the KDML105 rice seed produced in Khon Kaen (code D) could maintain seed vigor, even under ambient conditions.
Seed vigor classification using SV-RICE package
The SV-RICE package consists of 4 important steps, namely imaging, image processing, curve fitting and clustering. After curve fitting, radicle emergence indices, such as MaxRE, MRET, t50, U7525 and AUC, were calculated and these were used to classify rice seed vigor using a single index of radicle emergence via analysis of variance (ANOVA), at a significance level of p ≤ 0.05, followed by post hoc tests with Tukey’s honest significant difference (Tukey's HSD). For example, rice seed after 12 months of storage under controlled atmosphere conditions was classified into 2, 3 and 2 groups using MaxRE, MRET and t50, respectively, whilst seed after storing under ambient conditions for 12 months was classified into 5, 2 and 3 groups, respectively, using the same indices (Table 1). The results varied according to the radicle emergence index used.
Table 1
Radicle emergence indices for various varieties of Indica rice seed after 12 months of storage under different conditions using image processing assay in SV-RICE
Code | Controlled atmosphere storage* | Ambient storage† |
---|
Maximum radicle emergence (%) | Mean radicle emergence times (hours) | Radicle emergence speed (hours) | Uniformity of radicle emergence (hours) | Area under the curve of the radicle emergence fitted curve‡ | Maximum radicle emergence | Mean radicle emergence times (hours) | Radicle emergence speed (hours) | Uniformity of radicle emergence (hours) | Area under the curve of the radicle emergence fitted curve |
---|
A | 98.0 ab§ | 76.3 cd | 74.9 ab | 16.6 c | 206.6 ab | 72.3 c | 173.7 abc | 203.8 a | 108.3 a | 66.0 de |
B | 97.0 ab | 96.6 ab | 88.4 a | 53.4 ab | 191.9 cb | 48.0 d | 170.4 abc | 195.9 a | 105.3 a | 48.5 e |
C | 99.3 ab | 84.3 cb | 79.4 a | 34.3 bc | 208.2 ab | 98.8 a | 117.8 cd | 112.0bc | 48.9 bc | 177.2 b |
D | 99.5 ab | 82.9 bcd | 77.1 ab | 36.7 abc | 208.5 ab | 98.8 ab | 95.8 d | 89.2 bc | 44.4 bc | 200.3 ab |
E | 95.5 ab | 82.4 bcd | 78.2 ab | 30.9 bc | 195.8 b | 75.0 c | 178.0 ab | 198.4 a | 97.8 ab | 72.8 d |
F | 100.0 a | 63.9 d | 61.0 ab | 22.1 c | 230.2 a | 98.8 a | 77.0 d | 71.5 c | 34.6 c | 217.7 a |
G | 94.3 b | 109.9 a | 102.2 ab | 65.6 a | 166.6 c | 4.5 e | 171.5 a | 214.4 a | 113.6 a | 4.3 f |
H | 100.0 a | 85.1 cb | 77.7 b | 54.7 ab | 202.7 b | 93.5 b | 129.8 bcd | 133.4 b | 96.6 ab | 143.1 c |
Pr > F | 0.0060 | < .0001 | < .0001 | 0.0002 | < .0001 | < .0001 | < .0001 | < .0001 | < .0001 | < .0001 |
CV (%) | 5.701261 | 9.587863 | 8.658553 | 33.41275 | 5.650477 | 5.960406 | 17.63861 | 15.49259 | 28.06585 | 8.549533 |
* Seeds stored at 15°C and 37% RH |
† Seeds stored at 29°C and 53% RH |
‡ Area under the curve is the integration of the fitted curve between t = 0 and 300 hours |
§ Means within a column of each factor with the same lowercase letters are not significant at p ≤ 0.05 based on the Tukey’s honest significant difference test |
Generally, the SV-RICE package provided a systematic analysis of results, but it was slightly different from visual analysis (see Supplementary material 1). Differences between detection of radicle emergence based on visual analysis versus the SV-RICE package were analyzed using a box plot (Fig. 3) that showed there was an increase according to the germination time. Rice seeds were dispersed on the steel blue seed germination blotter paper using a mask to ensure an accurate and reproducible spacing. Therefore, clustering of seeds was prevented as much as possible. However, after 120 hours rice seedlings were touching each other and this caused clustering, resulting in greater differences between the visual analysis and the SV-RICE software. Interestingly, the SV-RICE package was especially optimized with Indica rice seed stored under a controlled atmosphere at 6 months (Fig. 3b). Nevertheless, the image processing in the SV-RICE package was effective when used with samples stored under ambient conditions for 12 months due to their low percentage radicle emergence (Fig. 3c).
The image processing algorithm of the SV-RICE package was effective in evaluating seed vigor for Indica rice grown in Thailand, although radicle emergence behavior differed markedly according to the conditions and storage time (Fig. 4). At 96 hours after the start of imbibition, the percentage radicle emergence of RD6 (Fig. 4e) was less than for KDML105 at the same time (Fig. 4b). The seed vigor of glutinous rice (code G) was lower than another variety (code A) because it of higher values for MRET, t50 and U7525 but lower values for MaxRE and AUC.
The SSAA test was correlated with a single radicle emergence index derived from the SV-RICE package (Fig. 5). Negative linear correlations were observed, with a high SSAA test result correlating with low values for MRET, t50 and U7525 and vice versa (p ≤ 0.05). Therefore, the classification of rice seed vigor using a single radicle emergence index derived from the SV-RICE package was likely to reflect the SSAA test. However, the use of multiple radicle emergence indices together from the SV-RICE package, would give more precise and accurate results. K-means clustering was used to partition the rice seed samples into K clusters in which each observation belonged to the cluster with the nearest mean using all the radicle emergence indices.
After clustering using the SV-RICE package, all the Indica rice seed samples used in the experiment were classified into 2 clusters: high and low seed vigor (Fig. 6e).
The optimal number of clusters can be adjusted as needed, although the appropriate number was recommended according to the silhouette method.
Figure
6d illustrates a k-mean cluster diagram for the 40 rice seed samples used in the experiment, which differed in genetics, production and storage time and storage condition. At 0 months of storage, old seed and glutinous rice seeds were classified as having low-vigor (Fig.
6a-c). Generally, the vigor of rice seed stored under controlled conditions was higher than under ambient conditions; the results for each sample are detailed in Supplementary material 2.
The SV-RICE clearly classified KDML105 rice seed from different production years, even though it could not to classify the samples from the different production areas (Fig. 7). KDML105 rice seed produced in 2017 was clearly distinguished from 2019 (Fig. 7c), although there was no apparent difference in the seeds coat based on observation with the naked eye (data not shown).