PosterPDF Available


Improving nutrient and water uptake in crops is one of the major challenges to sustain a fast-growing population that faces increasingly nutrient limited soils. Root hairs, which are specialized epidermal cells, compromise up to 70% of the total root surface area. Therefore, it is likely that root hairs are important for nutrient uptake from the soil. Microscopy provides a mean to record root hairs as digital images. However, quantifying root hairs in microscopy images remains a bottleneck because of a high degree of geometric complexity in their spatial arrangement. We present a methodology to automatically quantify phenotypic traits of root hairs in digital microscopy images. Our method uses a machine learning approach that is trained to classify root hair versus other root components and the image’s background. Image statistics of 20 training images are used to automatically classify root hair images taken with the same imaging protocol. Since root hairs may cross each other or form blobs of two or more hairs, we define metrics to distinguish these cases computationally. As a result, we measure the root hair traits number, length, density, diameter, orientation and overall complexity. We investigate root hair traits of rice, maize and common bean, under water and nitrogen stress. First tests with the new machine learning method suggest that our measured traits distinguish between genotypes and treatments in our study and pave the way towards identifying the genetic control of root hair traits.
Computational Methods
We us machine learning to classify root hairs, the lateral root from which root
hairs emerge and the image’s background.
Six training images were used to generate statistics to automatically classify all
root hair images taken with the same imaging protocol.
We developed first algorithms to measure length, density and diameter of
separate root hairs.
Improving nutrient and water uptake in resource limited soils is a major
challenge for agricultural research.
Root hairs are specialized epidermal cells that are important for nutrient
uptake from the soil by increasing the root-surface area.
Digital microscopy can record root hairs as images, but has never been used
for automated analysis.
We present amethodology to automatically quantify the geometric
complexity and spatial arrangement of root hairs.
Peter Pietrzyk
PhD student at the University of Georgia,
Dept. of Plant Biology, Bucksch Lab
Peter’s phenotyping interest is to develop imaging algorithms that characterize
the variability across multiple architectural scales in various crop roots.He has
a background in Aerospace Engineering ( and Geomatics (
Growth Conditions
Six commercial and traditional Thai rice varieties.
3-4 images were taken per variety.
Each variety was grown in a roll-up system using a germination paper soaked
with 0.5 mM calcium sulfate.
Samples for imaging were taken seven days after starting the roll-up.
Roots were preserved in 75% EtOH and stained with 0.25% toluidine blue for
collecting digital microscopy images.
Automated phenotyping of root hair traits
from microscopy images
Peter Pietrzyk1, Chartinun Chutoe2, Patompong Saengwilai2, Alexander Bucksch1,3,4
1Dept. of Plant Biology, UGA; 2Dept. of Biology, Faculty of Science, Mahidol University; 3Warnell School of Forestry and Natural Resources, UGA; 4Inst. of Bioinformatics, UGA
Improvement of our algorithm to distinguish intersecting root hairs or
clusters of two or more hairs.
Computing the traits root hair surface area, tip to main root distance, root
hair orientation provides ways to distinguish genotypes in full diversity
Extending our study to root hairs of maize and common bean is planned.
Modification of the experiment setup allows to determine influence of N/P/K
stress on root hairs.
Integration of our algorithms into DIRT (
Our results suggest that our algorithms distinguish the set of test genotypes
and potentially enable to identify genetic control of root hair traits.
Maledfai, Suphan Buri and Phitsanulok 2 were successfully distinguished.
KDML105, Lon and San Pa Tong are a suited test set for further improvement
of algorithm sensitivity.
Significant variation in root hair diameter was observed for Maledfai.
A bias is introduced to the results, because the current implementation is
sensitive to intersecting root hairs.
KDML105 Lon Maledfai San Pa Tong Phitsanulok 2Suphan Buri
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.