A noninvasive, machine learning–based method for monitoring anthocyanin accumulation in plants using digital color imaging

Abstract

Premise When plants are exposed to stress conditions, irreversible damage can occur, negatively impacting yields. It is therefore important to detect stress symptoms in plants, such as the accumulation of anthocyanin, as early as possible. Methods and Results Twenty‐two regression models in five color spaces were trained to develop a prediction model for plant anthocyanin levels from digital color imaging data. Of these, a quantile random forest regression model trained with standard red, green, blue (sRGB) color space data most accurately predicted the actual anthocyanin levels. This model was then used to noninvasively monitor the spatial and temporal accumulation of anthocyanin in Arabidopsis thaliana leaves. Conclusions The digital imaging–based nature of this protocol makes it a low‐cost and noninvasive method for the detection of plant stress. Applying a similar protocol to more economically viable crops could lead to the development of large‐scale, cost‐effective systems for monitoring plant health.

Seasonal pigment fluctuation in diploid and polyploid Arabidopsis revealed by machine learning-based phenotyping method PlantServation

“…6 and 7 ). Among model types, a random forest model, a machine learning method, performed the best, consistent with a previous study of pigment estimation from color in Arabidopsis in the laboratory 46 (see Supplementary Fig. 6 and Methods for details).…”

PlantServation: time-series phenotyping using machine learning revealed seasonal pigment fluctuation in diploid and polyploidArabidopsis

“…To estimate the anthocyanin content from the color information of images, we used another set of plants to collect color information from images and compared it with the actual pigment content measured experimentally. We regressed the relative anthocyanin content per leaf area on L*, a*, and b* using a random forest model that was previously found to be effective in pigment estimation from color in Arabidopsis in the laboratory 44 (Table S4). Among the three color features, a* contributed the most to the variation in anthocyanin, followed by b* and L* (importance: a* 56.4%, b* 22.4%, L* 21.2%).…”

“…We estimated the time series of leaf anthocyanin content from the color information. As in a previous study on Arabidopsis thaliana in the laboratory, a random forest model functioned effectively for our dataset in the estimation of anthocyanin 44 . With the strength of capturing nonlinearity in data, a random forest model can be effective for samples from fields where the environment is heterogeneous, causing noise in the color information.…”

Image-based phenotyping to estimate anthocyanin concentrations in lettuce

“…There are many successful examples of anthocyanin predictions based on hyperspectral imaging combined with machine learning techniques ( Chen et al., 2015 ; Askey et al., 2019 ; Simko, 2020 ; Cho et al., 2021 ; Kim et al., 2021 ). However, these machine learning models require hyperspectral imaging systems with similar wavelengths as used to develop the machine learning model.…”