Plant phenotyping

High Throughput Screening

High-throughput screening using LemnaTec scanalyzerHTS technology compared to manual measurements

Plant phenotyping is the comprehensive assessment of plant complex traits such as growth, development, tolerance, resistance, architecture, physiology, ecology, yield, and the basic measurement of individual quantitative parameters that form the basis for the more complex traits. Examples for such direct measurement parameters are image-based projected leaf area, chlorophyll fluorescence, stem diameter, plant height/ width, compactness, stress pigment concentration, tip burn, internode length, colour, leaf angle, leaf rolling, leaf elongation, seed number, seed size, tiller number, flowering time, germination time etc. After all, plant phenotyping has been performed by farmers and above all breeders for the last seven or more thousand years, essentially since the days humans started to carefully select grasses to increase yield or enhance other desirable traits. In the past, phenotyping was mostly based on experience and intuition, in a process where measurement and interpretation were not separated. This highly integrated approach compensated for the human deficiencies in performing reproducible, objective measurements and dealt with the individual subjectivity factor of the phenotyping person

Object Sum Area
Object Extend X/Y
Convex Hull Circumference
Caliper Length

Experimental Design

In this section we want to compare two traditional measurements, namely length of the longest leaf and fresh weight of the shoot to digital parameters and show correlations.  The data described here originates from an experiment performed on the PhenoFab system containing ten miscanthus genotypes, each in fivefold replicates. The images where acquired and the plants have been watered daily using a High throughput Scanalyzer 3D system and LemnaControl for 35 days. The imaging system was configures to take four lateral views rotated for 90 degrees each and a top view resulting in a total of 8,750 images. Genotypes were equally distributed between and within a total of four conveyor belts in one compartment. After 35 days the plants have been harvested and height as well as the fresh weight was measured and compared to results from the last images that have been taken the same day.

The image processing was performed on LemnaGrid and a data mining layout was created with LemnaMiner that will reduce the whole parameter set down to 5 parameters that are most likely to have a high correlation with the manually measured parameters. The parameters that have been extracted are explained below.

 

Object Sum Area - The number of all pixels that have been identified as part of the object

Object Extend X/Y - The width and height of the bounding box that surrounds the object

Convex Hull Circumference - The Length of the circumference of the Convex Hull. The convex is the smallest convex envelope that contains all pixels that have been identified as part of the plant. It can be visualized as the shape formed by a rubber band stretched around the plant

Caliper Length - The longest line that can be drawn between any of the points within the convex hull (similar to a maimum diameter of the convex hull)

Fresh Weight and Object Sum Area
Manual measured Length and Object Extend Y

The correlation between the manual measured parameters and the digital parameters can be seen in the table below. The correlation measurement found here described the mean squared error of the linear correlation between Fresh Weight and Length.

Object Sum Area

Object Extend X

Object Extend Y

Convex Hull Circumference

Caliper Length

FGW

95,63%

60,89%

86,52%

86,50%

86,55%

Length

77,82%

63,60%

98,57%

95,58%

97,98%

 

It can be seen that Fresh Weight and Object Sum Area show the best correlation while the manual measured Length correlates best to the Object Extend Y.