4. History
• Phase-contrast microscopy and time-lapse micro-cinematography were
used to observe rabbit oocyte fertilization and embryo development.
• The author provided detailed documentation on the timing of oocyte
penetration by spermatozoa, polar body extrusion, formation of pronuclei,
and cell divisions in early embryos.
• Embryos were held in a temperature-controlled perfusion chamber and
time-lapse pictures were taken at 16 frames per minute on 16mm film
under phase contrast.
• The author concluded that the sequence of events during fertilization and
early development in vitro is not different from in vivo.
1970 issue of Fertility and Sterility contains the article, “In Vitro
Fertilization of Rabbit Ova: Time Sequence of Events” by Brackett
5. CONTROVERSE
• Time-lapse imaging provides intermittent imaging of embryo development that
may aid in embryo selection.
• A recent manuscript analyzed the predictive value of 35 morphometric,
morphologic and morpho kinetic variables of embryo development and concluded
that only two parameters pronuclei (PN) position at time of PN juxtaposition and
the absence of multinucleated blastomeres at the 2-cell stage (MNB2cell), were
potentially associated with live birth.
Can novel early non-invasive biomarkers of embryo quality be identified with time-lapse imaging to predict live birth?
J Barberet, C Bruno, E Valot, C Antunes-Nunes, L Jonval, J Chammas, C Choux, P Ginod…
Human Reproduction, 2019•academic.oup.com
6. • AI analysis of time-lapse
morpho kinetic data
of in vitro embryo
development predicted live
birth in oocyte donation
cycles with more than 90%
accuracy
• AI analysis of time-lapse
embryo development
predicted that a selected
embryo would spontaneously
abort with 77% accuracy
Fertil Steril. 2019; 112: e1-e466
7. Currently, population health data often used to train models tend
to be unrepresentative of the entire population, with the data set
usually limited to a largely Caucasian population and one that is
homogenous in age, health, and socioeconomic status
Celi LA, Cellini J, Charpignon M-L, Dee EC, Dernoncourt F, Eber R, et al. Sources of bias in artificial intelligence that
perpetuate healthcare disparities—A global review. PLOS Digital Health 2022; e0000022.
8. Embryos from Mestizos develop faster and have a higher implantation
rate than those from Caucasian and Asian patients
P-182 The impact of ethnic differences on embryo morphokinetics and clinical outcomes: the importance of racial admixture
C Iaconelli, D Braga, A Setti, P Guilherme, A. Iaconelli, Jr., E Borges Junior
Human Reproduction, Volume 37, Issue Supplement_1, July 2022, deac107.175
9. Oocyte Quality
• convolutional neural networks (CNNs) and support vector machines (SVMs) to
analyse images of oocytes to predict their developmental potential. These
technologies are used mainly to identify oocytes with higher developmental
potential by analysing images of oocytes and identifying morphological features.
• Kanakasabapathy et al. developed a CNN trained to predict the likelihood of
fertilization based on static oocyte images. Similarly, SVMs have been used to
classify oocytes based on their morphological features
Kanakasabapathy M.K Bormann C.L. Thirumalaraju P. Banerjee R. Shafiee H.
Improving the performance of deep convolutional neural networks (CNN) in embryology using synthetic machine-
generated images Human Reprod. 2020; 35: I209
11. Sperm Selection
McCallum et al. trained an algorithm using acridine orange-stained sperm to
identify high DNA fragmentation to predict the acridine orange level based solely
on a brightfield image. This algorithm demonstrated a moderate ability to identify
sperm with differing levels of DNA fragmentation
McCallum C. Riordon J. Wang Y.
et al. Deep learning-based selection of human sperm with high DNA integrity.
Commun Biol. 2019; 2: 250
12. Embryo Assessment
“One of the most popular DL models that has seen numerous implementations in
the field of assisted reproduction is CNNs. Such systems are trained using large data
sets of static embryo images, where they learn the required parameters associated
with specific qualities over time.”
13. “Bormann et al. demonstrated the power of CNN-based frameworks by conducting
a study evaluating the consistency of 10 embryologists in performing routine
clinical tasks such as selecting embryos for biopsy, cryopreservation, or discard.
They directly compared the embryologists’ performance with the CNN-based
approach and demonstrated that the CNN-based framework far exceeded human
performance regarding consistency in embryo scoring and decision making. This
study highlighted the potential use of CNN-based frameworks for embryo analysis
in improving the overall quality of patient care in the field of assisted reproduction.”
Bormann C.L. Thirumalaraju P. Kanakasabapathy M.K. Kandula H.Souter I. Dimitriadis I. et al.
Consistency and objectivity of automated embryo assessments using deep neural networks.
Fertil Steril. 2020; 113: 781-787.e1
14. Ploidy Prediction
The KidscoreTM D5 algorithm as an additional tool to morphological assessment and PGT-A in embryo selection: a time-lapse study
Eduardo Gazzo,1 Fernando Peña,1 Federico Valdéz,2 Arturo Chung,1 Claudio Bonomini,1 Mario Ascenzo,1 Marcelo Velit,1 and Ernesto Escudero1
2020 Jan-Mar; 24(1): 55–60.doi: 10.5935/1518-0557.20190054 PMCID: PMC6993168PMID: 31608616
“Kato et al. demonstrated that the KIDScore of blastocysts on D5 and the Gardner
grade of blastocysts were associated significantly with euploidy, with a higher
KIDScore and Gardner grade associated with euploidy. In addition, the study found
that female age, the number of embryonic frozen days, and morphokinetic
characteristics were correlated negatively with euploidy, indicating that euploidy
decreased as these factors increased This study reinforced the foundation that by
combining clinical parameters and morphokinetic imaging data, AI can successfully
predict euploidy in a noninvasive manner.”
15. Embryo selection for transfer
“An Australian-US collaboration, developed an AI platform called
“Life Whisperer” that was trained using DL to predict pregnancy
outcome based on data from ground-truth clinical pregnancy outcomes.
The model achieved an accuracy rate of 64.3% and improved
predictions by 42% over embryologists’predictions alone. Importantly,
their model outperformed previous models because it was developed,
trained, and tested using known positive fetal heartbeat outcomes,
emphasizing the importance of predicting live births over implantation
potential. ”
VerMilyea M. Hall J.M.M. Diakiw S.M. Johnston A. Nguyen T. Perugini D.et al.
Development of an artificial intelligence-based assessment model for prediction of embryo viability using static images
captured by optical light microscopy during IVF.
Hum Reprod. 2020; 35: 770-784
16. Cell tracking and embryo witnessing
“a recent study reports an AI-based witnessing software using
predeveloped deep CNN models to predict blastocyst development at the
cleavage stage and classify blastocysts based on their developmental
quality
This system generates a unique identification score for each embryo
within a cohort, which is used to determine whether the embryos
originated from the same patient at a later time point.”
Hammer K.C. Jiang V.S.Kanakasabapathy M.K. Thirumalaraju P. Kandula H. Dimitriadis I. et al.
Using artificial intelligence to avoid human error in identifying embryos: a retrospective cohort study.
J Assist Reprod Genet. 2022; 39: 2343-2348
17. Automation of micromanipulation procedures
“jiang et al. reported using CNN models to train images of cleavage
stage embryos. The CNN models were trained and tested to evaluate
and classify images into 12 classifications, spaced 30° apart, to provide
an accurate location on the ZP to perform AH. The results of the study
showed that the CNN-AH model was able to identify correctly the
appropriate region to apply laser AH on the ZP with 99.41% accuracy”
Jiang V.S Kartik D. Thirumalaraju P. Kandula H. Kanakasabapathy M.K. Souter et al.
Advancements in the future of automating micromanipulation techniques in the IVF laboratory using deep convolutional neural networks.
J Assist Reprod Genet. 2023; 40: 251-257
18. Quality Management
“Cherouveim et al. aimed to use AI-based tools as quality assurance
measures to assess the performance of attending physician and
embryologists in several procedures, such as embryo transfer, embryo
vitrification, embryo warming, and trophectoderm biopsy. The AI
algorithm used in the study was a CNN that was developed and trained
as a supervised binary classifier to predict implantation outcome using
embryo images collected at 113 hours after insemination.”
Cherouveim P. Jiang V.S. Kanakasabapathy M.K. Thirumalaraju P. Souter I. Dimitriadis I. et al.
Quality assurance (QA) for monitoring the performance of assisted reproductive technology (ART) staff using Artificial intelligence (AI).
J Assist Reprod Genet. 2023; 40: 241-249
19. Finally…
New potential errors have been detailed in a 2019 analysis by Challen et al. One
example notes the potential for error resulting from discrepancies between the data
used to train AI systems and the real-world clinical scenario due to limited
availability of high-quality training data. AI systems are not as equipped as humans
to recognize when there is a relevant change in context or data that can impact the
validity of learned predictive assumptions. Therefore, AI systems may unknowingly
apply programmed methodology for assessment inappropriately, resulting in error.
Artificial intelligence, bias and clinical safety.
January 23, 2019
Challen R, Denny J, Pitt M, et al. BMJ Qual Saf. 2019;28(3):231-237.