Visualization of Root Growth

Ting-hsien Lin, Zhaohua Ding, Kikuo Fujimura, and Hideo Ishikawa

Department of Computer and Information Science and Biomedical Engineering Center and Department of Plant Biology Ohio State University, Columbus, OH 43210 In this project we attempt to visualize how root elongates by analyzing 2D time-varying series of images. It is of interest to plant biologists to find out how plant roots react to various environmental factors such as gravity and temperature. To find out active growing region within the root, marks are deposited on the surface of the root. A method is presented to identify fast-growing regions within roots by using a novel technique to measure `relative growth' for a moving point set.

Motivation

The project is being conducted as joint research between researchers in Dept. of Computer and Information Sceince and Dept. of Plant Biology. Biologists are interested in studying the effects of various environmental conditions (including light, gravity, and moisture) on the growth of the root of a plant. Some of the research questions we would like to study include:

Which part of the root grows most actively over time ?
How do the rates of growth of the outside/inside of the root compare?
How quickly does the root of the plant react when environmental conditions change?

Available Input

The picture taken in the lab shows a snapshot of a root grown in transparent gel. Depending on the experimental conditions and the expected rate of growth, images are grabbed and stored with inter-picture delays ranging from 1 minute to 30 minutes. See a sample growth movie. The root is laid horizontally in the beginning. The tip of the root begins to grow toward below as time passes. Small beads are deposited on the surface of the root before the experiment begins which can be used as landmarks to measure growth.

Current Approach and Preliminary results

The objective is to identify fast-growing regions within the root. Apparently, the tip of the root moves fastest. However, this does not necessarily indicate that the tip is the most fast-growing region, since the tip is pushed forward by the rest of the root. Thus, we need to identify `relative growth' of various parts of the root. To achieve the goal, we take following steps.

Segmentation of the images: We first identify features in the images, such as beads and the boundary of the root by using image processing techniques such as edge detection and border tracking. The following is a sample picture after edge detection.
Triangulation of feature points: Bead locations in each frame identified in the segmentation step are used to measure relative growth. We first triangulate the point set as below.
Visualization of the result: As a first step to measure relative growth, we need to establish point correspondence between frames. We use spatio-temporal coherence and root growth criteria such as possible growth rate to obtain the correspondence. Next, we compute the growth rate for each part (centering about each bead) of the root. The definition of root growth can be ambiguous. Consider two beads A and B in frame 1. Suppose the distance between A and B increased from frame 1 to frame 2. Which part of the root has grown, area near A or near B or somewhere inbetween?
We approach this problem by defining edge growth rate and point growth rate. The edge growth rate of an edge AB is the change of length of edge AB over time. The point growth rate of point A (bead A) is the weighted average of the edge growth rates of all edges connected to point A. The weights for averaging is different for horizontal growth (parallel to the spine of the root) and for vertical growth (perpendicular to the spine).
We compute the edge growth rates first and then the point growth rates at each bead point using edge growth rates. The point growth rate for one bead is a time-varying variable. The rate is plotted against time axis and then smoothed using a B-spline so as to supress temporal aliasing. To visualizing the growth rate of each point, we apply a technique similar to Gouraud smooth shading to the point growth rates, and finally the growth rate map is charted on a 2-dimensional time-varying images.
Below is a picture after visualization step. This picture indicates the growth rate parallel to the spine. The coloring scheme is shown at the left upper corner. We call it the "weather map" of the growth rate. We also include a weather map motion clip for horizontal (parallel to the spine) growth taken over a series of images. (The vertical motion clip is not included, although it is also available.) The motion clip illustrates how the `hot spot' changes over time as the root grows.

Summary

We have been developing a method for visualizing plant root growth based on the concept of relative growth. It makes it possible to characterize various root growth patterns automatically, which can serve as a basic block for builing a database for plant functions. We are currently pursuing a robust method for identifying growth patterns and a method for comparing growth patterns.

References

Barron, J.L. and A. Liptay. "Optical flow to measure minute increments in plant growth." Bioimaging. Vol. 2 (1994). pp 57-61.

Ishikawa, H., H.K. Hasenstein, and M.L. Evans. "Computer-based video digitizer a nalysis of surface extension in maize roots." Planta. Vol. 2 (1991). pp 3 81-390.