ABSTRACT
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, are important drivers of nutrient and water uptake from the soil.
Microscopy provides a mean to record root hairs as digital images.
However, due to their geometry and complex spatial arrangements
quantifying root hairs in microscopy images manually remains a
bottleneck. Manual selection of representative root hairs can result in
inaccurate estimations of root hair traits and misrepresentation of root
hair functions. We present a method to quantify phenotypes automatically
by measuring all individual root hairs in digital microscopy images. Our
method uses random forests classification to separate root hair from the
parent root and the image background. We define metrics to evaluate
segments of root hairs that intersect or form blobs of two or more root
hairs. Using simulated annealing for combinatorial optimization, we
reconstruct individual root hairs by resolving intersections in a
globally optimal way. As a result, we measure the root hair length, its
distribution, and root hair density in each image. We demonstrate our
method on examples of three maize cultivars under phosphorus, nitrogen,
and potassium stress. Results show that our measurements of root hair
traits strongly correlate with manually measured data in mean root hair
length (R 2: 0.72 to 0.85,p <.001), as well as in root hair density
(R 2: 0.38 to 0.66, p <.001). We
show that our method computes reliable estimates of root hair length,
density and their distributions along the root on complex root hair
arrangements in maize. We believe that our study paves a way towards
identifying the genetic control of root hair traits and increased
agricultural production.
Keywords: root hair, abiotic stress, phenotyping, machine
learning, simulated annealing, trait distribution
INTRODUCTION
Branching patterns in biology occur at all spatial scales, organismal
levels and for many physiological, protective or reproductive reasons
and are most prominent in the plant kingdom [1]. Hair-like
structures with a high length to width ratio are specific objects that
branch off and extend from the organism’s surface and have diverse
functions across biology [2]. To study their function it is
necessary to quantify their shape and arrangement accurately, which,
despite of their simple shape—in comparison to a multi-level branching
architecture—remains a challenge. Modern imaging tools can capture
digital images of these hair-like structures, but extracting all of them
individually from the image is ambiguous if they occlude each other
partially. The occlusion is especially prevalent in root hairs, which
are elongated epidermal cells extending from the root surface of a
plant. By increasing the root surface area and extending away from the
root surface into the soil, root hairs can increase water and nutrient
uptake from the soil [3].
We demonstrate an algorithm to resolve occlusion in 2D microscopy images
of root hairs. Researcher studying morphological traits of root hairs
traditionally use microscopy images and quantify root hair density and
length manually. Not only is this manual trait measurement of root hair
traits extremely tedious but the interpretation of 2D images
representing root hairs is also subjective. If measurements are done
automatically, only the total area or a profile of root hair length
along the root can be extracted [4, 5]. Other studies used X-ray
computed tomography (CT) to scan roots and root hairs in soil, but still
used manual tracing to segment root hairs from 3D X-ray CT scans
[6-8].
The challenge to extract individual root hairs from microscopy images is
that all intersections of root hairs must be resolved in the 2D
projection of the image. A single intersection of two root hairs can be
instantly resolved by determining the straightest solution from a small
number of possible combinations. By increasing the number of root hairs
and intersections, however, the number of potential combinations
increases in real scenarios to trillions of possible outcomes. We
present an approach to strategically resolve intersections and extract
individual root hairs in feasible computational time. As such, our
approach allows to measure root hair traits, like length and density,
and their distributions within a root sample at a much finer resolution
than previous 2D approaches.
MATERIAL AND METHODS