RSA extraction from 2D images for processing high-throughput data

Julien Diener (1, 2) , Philippe Nacry (3) , Bertrand Muller (4, 2) , Anne Diévart (5, 2, 6) , Yann Guédon (1, 2, 6) , Christophe Godin (1)

1 VIRTUAL PLANTS - Modeling plant morphogenesis at different scales, from genes to phenotype
2 UMR AGAP - Amélioration génétique et adaptation des plantes méditerranéennes et tropicales
3 BPMP - Biochimie et Physiologie Moléculaire des Plantes
4 LEPSE - Écophysiologie des Plantes sous Stress environnementaux
5 DAR - Développement Adaptatif du Riz [AGAP]
6 Cirad-BIOS - Département Systèmes Biologiques

Automatized acquisition systems of Root System Architecture (RSA) are now readily available for developmental research and provide high-throughput image data. Here we consider root system grown in petri plates. The analysis of these data (RSA) is currently a major challenge in understanding root development. However, suitable tools to process them efficiently are usually missing, thus reducing the possible scaling of processing and results. Existing tools either focus on specific applications, on simple structures (e.g. one root segment) or require long manual work. We present a processing pipeline that makes it possible to extract from an image the whole architecture of root systems, with minimal or no user intervention. In order to obtain this result, the problem was decomposed in several steps: filter and label the input image, extract the image skeleton as a general graph structure and then convert it into a tree structure representing the visualised RSA, using a priori knowledge to solve inconsistencies. The presented algorithm is inspired from publication of different scientific fields (blood vessel system extraction from 3D images, skeletization of virtual objects), but have been developed specifically for the extraction of the whole RSA from images acquired with traditional experimental protocol.

Mots clés: Root architecture; high-throughput; image analysis

Domaines: Modélisation et simulation; Traitement des images [eess.IV]; Biologie végétale

Julien Diener, Philippe Nacry, Bertrand Muller, Anne Diévart, Yann Guédon, et al.. RSA extraction from 2D images for processing high-throughput data. CSHL: Automated Imaging & High-Throughput Phenotyping, Apr 2012, Cold Spring Harbor, United States. 2012. ⟨hal-00915034⟩