O-005 Reducing inter-observer and intra-observer variability of embryo quality assessment using deep learning

Study question

Does deep learning for embryo quality assessment reduce inter-observer variability?

Summary answer

An AUC of 87.65% was obtained for predicting blastocyst quality with a deep-learning algorithm trained by five embryologists who had good agreement between themselves

What is known already

Time-Lapse (TL) allows continuous observation of embryo development in a controlled and stable environment. Recently the use of deep learning, in particular convolutional neural networks have been introduced to enhance blastocyst image classification using the growing TL image and video data.

Study design, size, duration

A total of 409 embryos (5 images per embryo for a total of 2 045 images) were included in this retrospective study between 2016 and 2020.

Participants/materials, setting, methods

A machine-learning algorithm (Retinanet) was trained to recognize 2 045 blastocyst images from 409 embryos on 2560x1928 images and output 500x500 images with the blastocyst centered on the image. Five embryologists classified the blastocysts using Gardner’s grading system. Each image was associated with one final grade using a majority voting system. The dataset was split into a training and validation set (1 640 images plus data augmentation) and a testing set (405 images).

Main results and the role of chance

Fair agreement was found between the 5 embryologists when grading the embryo using Gardner’s grading system, with a maximum weighted kappa score of 39.60% reached.

As for the intra-observer variability, we show that for the same embryologist grading the same embryo after a 3 month “wash out” period, in 12% of the cases the embryologist changes the grade and the fate of the embryo, meaning that an embryo that was transferred/frozen during the first annotation period was discarded during the second one, or an embryo that was discarded during the first annotation period was transferred/frozen during the second one.

An Area Under the Curve (AUC) of 87.65% was obtained when testing the quality of 81 embryos (405 images) after training our algorithm on 54 038 images.

For external validation we tested the algorithm with annotations of the test set from embryologists coming from another fertility center. An AUC of 82.72% was obtained.

Limitations, reasons for caution

The scarce number of images available in our training set compared with data sets from other more consequent clinics, and the fact that the algorithm was trained by embryologists does not suppress variability entirely. The GoogLeNet algorithm was not fined tune and was used as is.

Wider implications of the findings

AI is showing precious value the field of embryology, from enhancing blastocyst quality prediction to removing inter-observer subjectivity. A possible evolution to our framework would be to predict the Gardner’s grading system for each morphological parameter.

Trial registration number

not applicable

P–793 Validation of French in vitro fertilization (IVF) guideline during Covid–19 pandemic by the research of Sars-Cov–2 RNA in the follicular fluid (FF) after egg retrieval

Study question

Is it possible to find viral Sars-Cov–2 RNA in FF of women undergoing treatment during Covid–19 pandemic that may compromise gamete and embryo safety?

Summary answer

No viral RNA was detected in tested FF of women undergoing IVF in compliance with recommendations. This was reassuring and supported good medical practice.

What is known already

Risks due to SARS-CoV–2 during IVF remain difficult to assess despite the screening recommended by French health authorities based on a symptom questionnaire of the couple (systematic testing by RT-PCR for the virus before egg retrieval (ER) is not mandatory). In this context, this is a real challenge for IVF laboratory to guarantee procedure, patients, gametes and embryos safety. Most studies have reported the absence of virus in sperm. No data are available for FF and only one study looked for the presence of the virus in oocytes of Covid-affected patients (Barragan M et al, 2020).

Study design, size, duration

Between June 17 and September 24, 2020, FF of consenting women were prospectively collected and symptom questionnaire recorded. During this period, women undergoing IVF in our center did not benefit from systematic PCR testing for the virus within 72 hours prior to ER through our health authorities’ recommendations. All collected FF were retrospectively tested to research viral RNA by RT-PCR and patients were recalled to answer an epidemiological follow-up questionnaire.

Participants/materials, setting, methods

For all couples, symptom questionnaires were prospectively recorded and verified at each step of IVF procedure. For all consenting women, a sample of 1 ml of FF was collected the day of ER and stored at –80 °C. After thawing, a Sars-Cov2 multiplex RT-PCR using CFX96 (Biorad*) was performed, after RNA extraction using Nimbus (Seegene*). A comprehensive epidemiological evaluation was made afterwards by phone interview and data were recorded and analyzed.

Main results and the role of chance

A total of 183 women was included out of the 214 treated during this period (85.5%). Retrospective epidemiological evaluation showed that 8 patients contracted Covid more than 2 months before the ER, 6 more than 2 months after and only one patient 1 month after ER (diagnosis based on pathognomonic signs as agueusia and anosmia or/and positive PCR ). We observed a prevalence of symptomatic Covid forms in our IVF population of 8.2% during a 6-month period surrounding their IVF cycle. Moreover, until the introduction of systematic testing by RT-PCR for the virus before ER since the end of September 2020, 3 patients have been cancelled out of the 403 planned for positive PCR despite a negative questionnaire, which represents a prevalence of asymptomatic forms on the day of the ER at 0.7%. All the 183 FF tested did not reveal any viral RNA detection, which was reassuring concerning our medical practice and patient compliance and transparency. The absence of detected viral RNA may be due to several reasons: 1) women were not infected the day of ER 2) women had an asymptomatic form of the disease with low viral load 3) FF is not a virus reservoir.

Limitations, reasons for caution

Not all patients were included (85.5%). Post-diagnosis stays uncertain because PCR tests at the beginning of the epidemic were not mandatory and hardly available.

Wider implications of the findings: The absence of viral RNA in FF of women only screened through a symptom questionnaire is reassuring concerning the safety of IVF during Covid pandemic.

Trial registration number

Not applicable

Utilization of in vitro maturation in cases with a FSH receptor mutation

PURPOSE

To identify the FSH receptor (FSHR) variant and efficacy of in vitro maturation (IVM) in a 28-year-old woman with secondary amenorrhea, primary infertility, and ovarian resistance to FSH, and to analyze the genotype-to-phenotype relationship in cases of FSHR mutation for the development of an IVM algorithm for use in patients with gonadotropin resistance syndrome (GRS).

METHODS

Oocytes retrieved after menstruation induction with norethisterone, followed by daily estrogen and an ovulatory trigger, underwent IVM, ICSI, and culture in a time-lapse (TL) incubator. Embryo transfers were performed on day 2, and after thawing on day 5. Genes associated with disorders of sex development were sequenced for both the patient and her parents. All reported cases of FSHR mutation were analyzed to investigate genotype/phenotypic relationships.

RESULTS

After ovum pickup, seven of 16 oocytes matured and all fertilized. After unsuccessful day 2 transfer, our patient delivered with a thawed day 5 blastocyst, the sole embryo without abnormal TL phenotypes. Genetic analysis revealed a new composite heterozygous FSHR variant. Analysis of our patient case with published cases of GRS revealed associations among FSHR variant genotype, location on the FSHR, functionality of tested variants, and type of amenorrhea. An algorithm for application of IVM for GRS patients was developed.

CONCLUSIONS

We report two novel variants of the FSHR. Although IVM successfully matured some oocytes, only one resulted in an embryo with normal TL phenotypes. We recommend FSHR genetic testing in GRS patients, which will help guide their suitability for IVM.

A new mutation identified in SPATA16 in two globozoospermic patients

PURPOSE

The aim of this study is to identify potential genes involved in human globozoopsermia.

METHODS

Nineteen globozoospermic patients (previously screened for DPY19L2 mutations with no causative mutation) were recruited in this study and screened for mutations in genes implicated in human globozoospermia SPATA16 and PICK1. Using the candidate gene approach and the determination of Spata16 partners by Glutathione S-transferase (GST) pull-down four genes were also selected and screened for mutations.

RESULTS

We identified a novel mutation of SPATA16: deletion of 22.6 Kb encompassing the first coding exon in two unrelated Tunisian patients who presented the same deletion breakpoints. The two patients shared the same haplotype, suggesting a possible ancestral founder effect for this new deletion. Four genes were selected using the candidate gene approach and the GST pull-down (GOPC, PICK1, AGFG1 and IRGC) and were screened for mutation, but no variation was identified.

CONCLUSIONS

The present study confirms the pathogenicity of the SPATA16 mutations. The fact that no variation was detected in the coding sequence of AFGF1, GOPC, PICK1 and IRGC does not mean that they are not involved in human globozoospermia. A larger globozoospermic cohort must be studied in order to accelerate the process of identifying new genes involved in such phenotypes. Until sufficient numbers of patients have been screened, AFGF1, GOPC, PICK1 and IRGC should still be considered as candidate genes.

DPY19L2 deletion as a major cause of globozoospermia

Globozoospermia, characterized by round-headed spermatozoa, is a rare (< 0.1% in male infertile patients) and severe teratozoospermia consisting primarily of spermatozoa lacking an acrosome. Studying a Jordanian consanguineous family in which five brothers were diagnosed with complete globozoospermia, we showed that the four out of five analyzed infertile brothers carried a homozygous deletion of 200 kb on chromosome 12 encompassing only DPY19L2. Very similar deletions were found in three additional unrelated patients, suggesting that DPY19L2 deletion is a major cause of globozoospermia, given that 19% (4 of 21) of the analyzed patients had such deletion. The deletion is most probably due to a nonallelic homologous recombination (NAHR), because the gene is surrounded by two low copy repeats (LCRs). We found DPY19L2 deletion in patients from three different origins and two different breakpoints, strongly suggesting that the deletion results from recurrent events linked to the specific architectural feature of this locus rather than from a founder effect, without fully excluding a recent founder effect. DPY19L2 is associated with a complete form of globozoospermia, as is the case for the first two genes found to be associated with globozoospermia, SPATA16 or PICK1. However, in contrast to SPATA16, for which no pregnancy was reported, pregnancies were achieved, via intracytoplasmic sperm injection, for two patients with DPY19L2 deletion, who then fathered three children.

Choanal atresia and hypothelia following methimazole exposure in utero: a second report

We report on a 3-year-old boy with bilateral choanal atresia, hypoplastic nipples, and developmental delay who had been exposed to carbimazole in utero because of maternal Graves disease. His combination of abnormalities and facial appearance strongly resembles that of a previously reported child exposed to methimazole (which is the active metabolite of carbimazole) in utero. We suggest that this represents a rare but distinct syndrome of methimazole teratogenicity, probably related to first-trimester exposure. Recognition of such teratogenic effects is clearly important for genetic counselling and for management of subsequent pregnancies.

Separation of the proteins of cerebrospinal fluid using gel electrofocusing followed by electrophoresis

The technique previously applied to serum can, with minor modifications, be applied to cerebrospinal fluid. Good protein patterns have been obtained and similarities as well as differences have been demonstrated between cerebrospinal fluid and serum. The patterns obtained should prove useful in neurological diagnosis.