Objective Definition of Rosette Shape Variation Using a Combined Computer Vision and Data Mining Approach Anyela Camargo1,Dimitra Papadopoulou2,Zoi Spyropoulou2,Konstantinos Vlachonasios2,

John H.Doonan1*,Alan P.Gay1*

1Institute of Biological,Environmental and Rural Sciences,Aberystwyth University,Gogerddan,Aberystwyth,Ceredigion,United Kingdom,2Aristotle University of Thessaloniki,Faculty of Science,School of Biology,Department of Botany,Thessaloniki,Greece

Abstract

Computer-vision based measurements of phenotypic variation have implications for crop improvement and food security because they are intrinsically objective.It should be possible therefore to use such approaches to select robust genotypes.

However,plants are morphologically complex and identification of meaningful traits from automatically acquired image data is not straightforward.Bespoke algorithms can be designed to capture and/or quantitate specific features but this approach is inflexible and is not generally applicable to a wide range of traits.In this paper,we have used industry-standard computer vision techniques to extract a wide range of features from images of genetically diverse Arabidopsis rosettes growing under non-stimulated conditions,and then used statistical analysis to identify those features that provide good discrimination between ecotypes.This analysis indicates that almost all the observed shape variation can be described by5 principal components.We describe an easily implemented pipeline including image segmentation,feature extraction and statistical analysis.This pipeline provides a cost-effective and inherently scalable method to parameterise and analyse variation in rosette shape.The acquisition of images does not require any specialised equipment and the computer routines for image processing and data analysis have been implemented using open source software.Source code for data analysis is written using the R package.The equations to calculate image descriptors have been also provided.

Citation:Camargo A,Papadopoulou D,Spyropoulou Z,Vlachonasios K,Doonan JH,et al.(2014)Objective Definition of Rosette Shape Variation Using a Combined Computer Vision and Data Mining Approach.PLoS ONE9(5):e96889.doi:10.1371/journal.pone.0096889

Editor:Hector Candela,Universidad Miguel Herna′ndez de Elche,Spain

Received August30,2013;Accepted April13,2014;Published May7,2014

Copyright:?2014Camargo et al.This is an open-access article distributed under the terms of the Creative Commons Attribution License,which permits unrestricted use,distribution,and reproduction in any medium,provided the original author and source are credited.

Funding:The work was funded by the Biological and Biosciences Research Council and the Welsh Government.The funders had no role in study design,data collection and analysis,decision to publish,or preparation of the manuscript.

Competing Interests:The authors have declared that no competing interests exist.

*E-mail:john.doonan@http://www.doczj.com/doc/b8aaf114e518964bcf847ca7.html(JHD);abg@http://www.doczj.com/doc/b8aaf114e518964bcf847ca7.html(APG)

Introduction

The goal of this study was to use a computer vision and data mining approach to compare the rosette shapes of the founders of a Multiparent Advanced Generation Inter-Cross(MAGIC) population.A genetic analysis of the same population was performed and reported previously[1].Objective computer-aided phenotyping has been proposed as a solution to the genotype-phenotype bottleneck[2],but there remain numerous technical challenges with regard to its implementation at the whole organism level.However,there has been little exploration of the ability of computer vision techniques to define and discriminate between phenotypes.

We chose Arabidopsis rosettes as our experimental material for three main reasons.First,the rosettes in this species(under our growth conditions)grow close to the ground and can be treated essentially as2-D objects,simplifying image acquisition and processing.Second,previous studies indicate that there is significant shape variation between accessions.Natural variation in continuously varying traits has been shown for morphological traits and for responses to stimuli.Examples of the former are morphological comparisons during development between Ler-0, Col-0and Ws-0ecotypes[3],quantitative trait loci(QTL)analysis of leaf and floral organ size of162recombinant inbred lines(RIL) from a reciprocal cross between Ler and Cvi[4]and seed size of the iku2-1,fis2-1,arf2,pAP1::ARF2mutants and Col-0and Ler-0 ecotypes[5].Examples of the latter are the effects of drought,low temperature and differing levels of UV-B on chlorophyll-fluores-cence on growth[6]and the natural variability of23accessions in response to nitrogen[7].Third,Arabidopsis is a widely used model system with sophisticated genetic and genomic resources[8] available for dissecting biological processes.Forward genetic approaches have been used to study mutants with strong phenotypic effects providing insight into the underlying molecular functions.While this approach is extremely useful as a research tool,commercial plant breeding often requires exploitation of continuous variation.Analysis of continuous variation in breeding populations is more demanding,but effective automation of the phenotype measurements would have huge advantages for crop improvement and food security.Arabidopsis is also a good model for studying continuous variation,with the advantage of thoroughly investigated genomics[9,10].The native range is North-Western Eurasia and it has recently colonised other parts of the world during the Columbian Exchange[11].Local populations have often diverged,to a degree depending on factors such as time of separation and differential selection.The species is therefore a useful model to study natural variation,its underlying genetic basis and its consequences.

Natural variants provide material for studying genome evolu-tion and the genetic dissection of complex traits.The Arabidopsis