In-vivo Starch Quantification:

Spatial Mapping via X-ray MicroCT Imaging and Machine Learning

Starch is the central energy storage molecule used by plants to fuel respiration and growth during periods of limited or no photosynthesis.  Recently, the relative starch concentration of plants entering drought was linked to mortality probability, as the relative pools of starch and other non-structural carbohydrates are necessary for maintaining cellular functions, producing chemical defense compounds to prevent biotic attack, and regrowing foliage. Yet, no techniques exist for in vivo observation and quantification of non-structural carbohydrates. We developed a novel technique that uses machine learning and x-ray microCT imaging (Fig. 1) to accurately quantify and spatially locate non-structural carbohydrates in the vascular tissue of grapevine plants. To do this, we first analyzed the data structure of X-ray absorption values within the stem identifying the variance as a key property relating to cellular regions that appeared as ‘full’ or ’empty’ of starch (Fig. 2). Next, using the variance map, along with other image filters, we trained a WEKA machine learning algorithm to identify regions where parenchyma cells appeared full or empty of starch (Fig. 3). We then developed a relationship between machine-classified regions as ‘full’ or ’empty’ of starch and enzymatically-assayed non-structural carbohydrate concentration (Fig. 4). Based on this relationship, we then spatially mapped patterns of non-structural carbohydrate depletion at micrometer resolution in vivo for the first time (Fig. 5). We anticipated that such high spatial-resolution in vivo starch monitoring should enable novel research directions across the plant sciences.


relevant publications

Earles, J.M.*, Knipfer, T.K.*, Tixier, A., Reyes, C., Orozco, J., Zwieniecki, M.A., Brodersen, C.R., and McElrone, A.J. (in review). In-vivo quantification of starch reserves in plants using X-ray microCT imaging and machine learning. *Authors contributed equally



Craig Brodersen – Yale University
Thorsten Knipfer – University of California, Davis
Andrew McElrone – USDA Agricultural Research Service
Jessica Orozco – University of California, Davis
Clarissa Reyes – University of California, Davis
Aude Tixier – University of California, Davis
Maciej Zwieniecki – University of California, Davis