Toward a standardised annotation of morphokinetical parameters for an automatic early prediction of the in vitro development potential of bovine embryos
Résumé
In a previous work, we proposed an algorithm to select a subset of discriminant morphokinetical parameters within a larger predefined set (116 parameters) and to predict 6 major bovine embryo profiles of in vitro development. The algorithm relies on flexible combinations of the discriminant parameters selected within the four first cleavage cycles. The retained profiles were arrested embryos (embryos without mitotic activity, showing signs of life); dead embryos (embryos with all cells dead); anarchic embryos (embryos with abnormal morphological and/or kinetical development: some of these embryos can result in a blastocyst); not hatched blastocysts (blastocysts not hatching by 8 dpi); hatching blastocysts (blastocysts hatching in vitro from 7.3 to 8 dpi); and early hatching blastocysts (blastocysts hatching from 6 to 7.2 dpi) (Reis et al. 2018 Anim. Reprod.). The aim of the present work was to develop a ready-to-use software (EasyPickAndPredict) allowing the extension of this methodology to other embryologists. The software of high portability was built with the JAVA language and embedded with the classifier (R language, R Foundation for Statistical Computing, Vienna, Austria) and a user’s help for annotation (Adobe Premiere Pro CS5, San Jose, CA, USA). The predictive software is easy to handle, fast to load, and has high portability. The “manual annotation” function is based on click actions to annotate the discriminant parameters. The “prediction” function calls the embedded classifier. The “report” function creates customised reports including the embryo classification, the summary of the measures, and the accuracy of the prediction (vote system). The “help” function calls an audiovisual guideline with annotation specifications for all the morphokinetic parameters currently described (including the discriminant subset of parameters). This document includes help to produce a good time-lapse video (16 min 51 s); annotation specifications for the cleavage cycles 1 to 4 (36 min 26 s); and a summary with examples of the 6 major embryonic morphokinetical profiles (16 min 47 s). The predictive graphical interface is easy to manipulate and automatically extracts the value of each discriminant parameter on the time-lapse picture as validated by the user click. The functions “manual annotation”, “automatic prediction”, “help”, and “report” supply embryologists with a standard approach to predict and analyse morphokinetical profiles of their embryos. This standardised approach is necessary to improve the capacity of comparison of morphokinetical works in different laboratories and enhance knowledge about in vitro-produced bovine embryos.