Phenotypic characterization of floral trichomes of Cannabis sativa L. using computer vision and deep learning

Cannabis (Cannabis sativa L.) is cultivated by licensed producers in Canada for medicinal and recreational uses. The recent legalization of this plant in 2018 has resulted in rapid expansion of the industry, with greenhouse production representing the most common method of cultivation. Female cannabis plants produce inflorescences that contain bracts densely covered by glandular trichomes, which synthesize a range of commercially important cannabinoids (e.g., THC, CBD) as well as terpenes. Cannabinoid content and quality varies over the 8-week flowering period to such an extent that the time of harvest can significantly impact product quality. Cannabis flower maturation is accompanied by a transition in the color of trichome heads that progresses from clear, to milky, to brown (amber) and can be seen using low magnification. However, the importance of this transition as it impacts quality and describes maturity has never been investigated. We first collected a macro-photography dataset based on 4 commercially grown cannabis strains, namely 'Afghan Kush', 'Green Death Bubba', 'Pink Kush', and 'White Rhino'. Images were annotated with total THC potency measurements via HPLC analysis of whole flowers. To establish a relationship between trichome maturation and trichome head appearance changes (phenotype), we developed a novel automatic trichome gland analysis pipeline using deep learning. The pipeline, Automatic Trichome Analysis, first uses deep learning to localize and segment trichomes in images, followed by a KNN classifier injected with the clear-milky-brown heuristic as classes to quantify trichome phenotype progression during the 8-week flowering period. A series of clear, milky, and brown phenotype curves were recorded for each strain over the flowering period that were validated as indicators of trichome maturation and corresponded to previously described parameters of trichome development, such as trichome gland head diameter and stalk elongation. We also derived morphological metrics describing trichome gland geometry from deep learning segmentation predictions that profiled trichome maturation over the flowering period and were correlated with cannabinoid potency. We observed that mature and senescing trichomes displayed fluorescent properties that were reflected in the clear, milky, brown phenotypes. We also investigated potency prediction from images of cannabis flowers, and present a novel mosaic augmentation strategy using trichome heads that improves the model convergence of a neural regressor when compared to mixup and standard augmentations. We also found that the clear-milky-brown phenotype was correlated with cannabinoid potency, but ultimately these phenotypes were proxy labels for trichome volume, and thus extracting trichome volume from trichome head segmentations directly resulted in a more powerful correlation. Our results indicate the feasibility of automated trichome analysis as a method to evaluate the maturation of female flowers cultivated in a highly variable environment, regardless of strain. These findings have broad applicability in a growing cannabis industry where flower quality is a foremost at-tribute for medicinal and recreational uses.