03183nam a2200253 a 450000100080000000500110000800800410001910000150006024501700007526001460024552023130039165000320270465000100273665300160274665300350276265300170279765300100281470000180282470000160284270000160285870000210287470000160289570000180291110598242019-08-14       2018    bl uuuu     u00u1 u    #d1 aPEZARD, J.  aAdapting automated image analysis to breeding programs constraints for the characterization of the resistance to leaf rust and other diseases.h[electronic resource]  aIn: Proceedings of the International Cereal Rusts and Powdery Mildew Conference (ICRPMC):  Skukuza, South Africa, 23-26 september 2018.c2018  aDescription: Disease phenotyping methods used in breeding programs to characterize the level of resistance of breeding materials usually consist on visual scores (VS) of disease symptoms determined in field trials. VS are considered as high time-consuming and rely on experienced operators. Nevertheless, up to date, it is the only method that has an efficient time/effort relationship considering breeding constrains. The objective was to develop a phenotyping methodology based on automated image analysis (AIA) for leaf diseases, adapted to the constraints of a breeding program. 410 wheat lines from 5 different breeding programs were sowed in three field trials, as part of the materials tested in 2017 at the multi-disease phenotyping platform INIA-CIMMYT, Uruguay. One trial was inoculated with Puccinia triticina isolates the second with Zymoseptoria tritici isolates and the third had natural infection of P. striiformis f. sp. tritici. Six flag leaves per genotype were cut and scanned with a flatbed scanner. A script was developed in the ImageJ software to autonomously recognize and measure the leaf diseased surface. Disease recognition and surface measurements were based on the different threshold color patterns of each disease. Host response was also determined for leaf and stripe rust, measuring the ratio of necrosis-chlorosis/sporulation area of lesions. AIA recognized the different diseases (error<5%). The diseased surfaces obtained by AIA correlated significantly and positively with the VS measured for the three diseases. Host responses estimated by AIA were the same as determined visually, (error<5%). AIA was fast, a mean of 214 leaves/hour analyzed, taking into account the adjustments of color thresholds and the validation of AIA. However, the time to prepare and scan the leaves was higher than the VS: a mean of 205 lines could be scanned per person/day while a mean of 402 lines per person/day could be visually scored. Adjustments to the scan methodology are being carried out to enhance the speed at this step. Nevertheless, AIA can be a performing alternative to VS in limited panels or mapping populations that undergo QTL analysis, where precise measurements of quantitative resistance variables are required to detect QTL with moderate effects and QTL interactions.  aENFERMEDADES DE LAS PLANTAS  aTrigo  aINIA-CIMMYT  aPLATAFORMA FENOTIPADO DE TRIGO  aRUST DISEASE  aWHEAT1 aFERNANDEZ, P.1 aPEREYRA, S.1 aQUINCKE, M.1 aSAINT-PIERRE, C.1 aSINGH, P.K.1 aAZZIMONTI, G.