Submitted to: IEEE Signal Processing Magazine
Publication Type: Peer reviewed journal
Publication Acceptance Date: 5/1/2007
Publication Date: 5/15/2007
Publication URL: http://ieeexplore.ieee.org/xpls/abs_alll.jsp?isnumber=4202144&arnumber=4205096&count=33&index=21
Citation: Shyu, C., Green, J.M., Lun, D.P., Kazic, T., Schaeffer, M.L., Coe, E. 2007. Impage Analysis for Mapping Immeasurable Phenotypes in Maize. IEEE Signal Processing Magazine. 24(3):116-119. Interpretive Summary: Computer-based phenotype (outward appearance) image analysis will aide bio-informaticians in the analysis of the ever-increasing gene sequence data, discover valuable knowledge in maize biology and related plant development, and understand subtle variations among different phenotypes. Furthermore, successful quantitative measurement of visual phenotypes will advance plant research by aiding the quest for the genes and/or environmental factors that cause a given visual phenotype. This paper introduces the field of plant genetics to the computer scientists engaged in image signal processing, discusses the challenges involved, and presents an image analysis system for precisely quantifying and mapping what is up until now qualitative (immeasurable) phenotypes in maize. This model could also be extended to other species within and outside the plant kingdom and may provide opportunity for cross-species phenotypic comparison at a more generic feature and gene level.
Technical Abstract: A majority of phenotypic variance in maize has qualitative aspects that are immeasurable by rulers or scalars. Image analysis may improve the phenotypic quantification by increasing the objectivity and granularity of quantification, which in turn may result in an increase in the rate at which the genes controlling phenotypic traits are isolated. This report describes the analysis of visible phenotypes for maize ears and leaves, using digitized images. This system handles images acquired in corn fields which, in contrast to images acquired in a greenhouse, present many processing challenges that stem from changing weather factors, shifting light and sub-optimal camera adjustments. The images acquired are first pre-processed and segmented. Then features are extracted and indexed in a high-dimensional space. Cluster indexed feature vectors have been used to score a disease, Southern Leaf Blight, characterize lesion mutants, and localize the captured phenotypes onto a genetic or physical map.