72 Evaluation of digital images of bovine embryos

Abstract

Bovine embryo transfer is an assisted reproductive technology that enables cattle producers to reach their reproductive, genetic, and financial goals. Proper evaluation of embryo quality and developmental stage is key to the success of the embryo transfer industry and is often acquired “on the job” as there are few embryo evaluation training opportunities. The goals of this project were to develop a bank of digital images of in vivo-produced bovine embryos, and to evaluate interrater reliability and variation of grade and stage scores and the assessment of the integrity of the zona pellucida (ZP) using an online platform. Frozen/thawed bovine embryos (Day 7) were placed in a commercially available holding medium, and images were captured using a high-resolution camera (112× magnification). Survey participants (n = 89) responded to questions on demographic and embryo evaluation experience and assigned grades and stages and evaluated the integrity of the ZP in 58 embryo images (the IETS embryo evaluation rubric was provided). The variation among raters within and between demographic categories was analysed using the “kappaetc” package with weights for ordinal variables (grade and stage) in Stata version 14.2 (StataCorp). The Kruskal–Wallis test was used to assess whether variation [measured as interquartile range (IQR)] in grade and stage ratings within individual were significantly different (P = 0.05) by demographic factors. Overall, interrater reliability (measured by Cohen’s Kappa, κ) for stage was moderate (κ = 0.57), whereas identification of grade and the ZP was substantial (κ = 0.69 and 0.73, respectively). Interrater reliability for grade was significantly lower among those with ≤2 years (n = 17) of experience (κ = 0.44) compared with those having 3–9 years (n = 29; κ = 0.57) and ≥10 years (n = 43; κ = 0.64). Similarly, interrater agreement for stage was significantly lower for those with ≤2 years of experience (κ = 0.58) compared with those having 3–9 years (κ = 0.68) and ≥10 years (κ = 0.75). Interrater agreement for stage was higher for those who did not use the IETS rubric compared with those who did (κ = 0.73 vs. κ = 0.63, respectively), and the variability was highest for Stage 6 and lowest for Stage 4 embryos. No demographic variables were significantly associated with interrater reliability for the integrity of the ZP. The variability of stage ratings (IQR) within individual significantly increased with more years of experience, with greatest IQR among those with ≥10 years of experience. The IQR within individual for grade decreased with increasing length of time rating, with the lowest IQR for the longest duration spent on rating the embryos. Overall, there was a moderate to substantial interrater agreement in embryo evaluation, which increased with experience. Further, evaluation of bovine embryo digital images may be a useful training tool instead of or in conjunction with a “hands-on” experience. Future surveys will involve multiple images of the same embryo and video to enhance the evaluation process and increase interrater reliability.