2025 Volume 27 Issue 1 Pages 7-18
Phenotyping is essential for crop breeding but is often time-consuming and labor-intensive, necessitating the improvement and efficiency. Hyperspectral (HS) imaging is a promising technology because it allows non-destructively analyze the internal state of plants and visualize the distribution of specific information. This study aimed to leverage this method to enable rapid and simple analysis of the chemical and physiological characteristics of plants. Specifically, we developed a non-destructive evaluation system for photosynthetic product contents in rice seedlings. Seedlings of the Japanese rice cultivar ‘Koshihikari’ and the Indian rice cultivar ‘IR64’ were grown under various light conditions, and HS reflectance data were extracted from images by an HS camera. After imaging, the rice seedlings were collected by parts, and the content of photosynthetic products (sucrose and starch) was determined by LC-MS analysis. Data from 198 leaf blades and 198 leaf sheaths were divided into training and validation datasets. Using HS reflectance data as explanatory variables and sucrose and starch content as response variables, we created a prediction model by partial least squares regression for each plant part and calculated the coefficients of determination (R2). Applying these prediction models for sucrose content (leaf blades: R2 = 0.58, leaf sheaths: R2 = 0.73) and starch content (leaf blades: R2 = 0.89, leaf sheaths: R2 = 0.68) to time-series HS reflectance data of the same plant allowed us to estimate the distribution of sucrose and starch content at the pixel level, visualizing their gradual increase under continuous light conditions. The measured values indicated an earlier increase in leaf blades followed by a delayed increase in leaf sheaths, this trend was also confirmed in the visualization images. These results suggest that HS imaging enables non-destructive visualization of differences and changes in photosynthetic product content at each part level in rice seedlings.
