Development of an artificial intelligence model for predicting the likelihood of human embryo euploidy based on blastocyst images from multiple imaging systems during IVF.

Can an artificial intelligence (AI) model predict human embryo ploidy status using static images captured by optical light microscopy? Results demonstrated predictive accuracy for embryo euploidy and showed a significant correlation between AI score and euploidy rate, based on assessment of images of blastocysts at Day 5 after IVF. Euploid embryos displaying the normal human chromosomal complement of 46 chromosomes are preferentially selected for transfer over aneuploid embryos (abnormal complement), as they are associated with improved clinical outcomes. Currently, evaluation of embryo genetic status is most commonly performed by preimplantation genetic testing for aneuploidy (PGT-A), which involves embryo biopsy and genetic testing. The potential for embryo damage during biopsy, and the non-uniform nature of aneuploid cells in mosaic embryos, has prompted investigation of additional, non-invasive, whole embryo methods for evaluation of embryo genetic status. A total of 15 192 blastocyst-stage embryo images with associated clinical outcomes were provided by 10 different IVF clinics in the USA, India, Spain and Malaysia. The majority of data were retrospective, with two additional prospectively collected blind datasets provided by IVF clinics using the genetics AI model in clinical practice. Of these images, a total of 5050 images of embryos on Day 5 of in vitro culture were used for the development of the AI model. These Day 5 images were provided for 2438 consecutively treated women who had undergone IVF procedures in the USA between 2011 and 2020. The remaining images were used for evaluation of performance in different settings, or otherwise excluded for not matching the inclusion criteria. The genetics AI model was trained using static 2-dimensional optical light microscope images of Day 5 blastocysts with linked genetic metadata obtained from PGT-A. The endpoint was ploidy status (euploid or aneuploid) based on PGT-A results. Predictive accuracy was determined by evaluating sensitivity (correct prediction of euploid), specificity (correct prediction of aneuploid) and overall accuracy. The Matthew correlation coefficient and receiver-operating characteristic curves and precision-recall curves (including AUC values), were also determined. Performance was also evaluated using correlation analyses and simulated cohort studies to evaluate ranking ability for euploid enrichment. Overall accuracy for the prediction of euploidy on a blind test dataset was 65.3%, with a sensitivity of 74.6%. When the blind test dataset was cleansed of poor quality and mislabeled images, overall accuracy increased to 77.4%. This performance may be relevant to clinical situations where confounding factors, such as variability in PGT-A testing, have been accounted for. There was a significant positive correlation between AI score and the proportion of euploid embryos, with very high scoring embryos (9.0-10.0) twice as likely to be euploid than the lowest-scoring embryos (0.0-2.4). When using the genetics AI model to rank embryos in a cohort, the probability of the top-ranked embryo being euploid was 82.4%, which was 26.4% more effective than using random ranking, and ∼13-19% more effective than using the Gardner score. The probability increased to 97.0% when considering the likelihood of one of the top two ranked embryos being euploid, and the probability of both top two ranked embryos being euploid was 66.4%. Additional analyses showed that the AI model generalized well to different patient demographics and could also be used for the evaluation of Day 6 embryos and for images taken using multiple time-lapse systems. Results suggested that the AI model could potentially be used to differentiate mosaic embryos based on the level of mosaicism. While the current investigation was performed using both retrospectively and prospectively collected data, it will be important to continue to evaluate real-world use of the genetics AI model. The endpoint described was euploidy based on the clinical outcome of PGT-A results only, so predictive accuracy for genetic status in utero or at birth was not evaluated. Rebiopsy studies of embryos using a range of PGT-A methods indicated a degree of variability in PGT-A results, which must be considered when interpreting the performance of the AI model. These findings collectively support the use of this genetics AI model for the evaluation of embryo ploidy status in a clinical setting. Results can be used to aid in prioritizing and enriching for embryos that are likely to be euploid for multiple clinical purposes, including selection for transfer in the absence of alternative genetic testing methods, selection for cryopreservation for future use or selection for further confirmatory PGT-A testing, as required. Life Whisperer Diagnostics is a wholly owned subsidiary of the parent company, Presagen Holdings Pty Ltd. Funding for the study was provided by Presagen with grant funding received from the South Australian Government: Research, Commercialisation, and Startup Fund (RCSF). 'In kind' support and embryology expertise to guide algorithm development were provided by Ovation Fertility. 'In kind' support in terms of computational resources provided through the Amazon Web Services (AWS) Activate Program. J.M.M.H., D.P. and M.P. are co-owners of Life Whisperer and Presagen. S.M.D., M.A.D. and T.V.N. are employees or former employees of Life Whisperer. S.M.D, J.M.M.H, M.A.D, T.V.N., D.P. and M.P. are listed as inventors of patents relating to this work, and also have stock options in the parent company Presagen. M.V. sits on the advisory board for the global distributor of the technology described in this study and also received support for attending meetings. N/A.

Diakiw SM ，Hall JMM ，VerMilyea MD ，Amin J ，Aizpurua J ，Giardini L ，Briones YG ，Lim AYX ，Dakka MA ，Nguyen TV ，Perugini D ，Perugini M ... -  《-》

Development of an artificial intelligence-based assessment model for prediction of embryo viability using static images captured by optical light microscopy during IVF.

Can an artificial intelligence (AI)-based model predict human embryo viability using images captured by optical light microscopy? We have combined computer vision image processing methods and deep learning techniques to create the non-invasive Life Whisperer AI model for robust prediction of embryo viability, as measured by clinical pregnancy outcome, using single static images of Day 5 blastocysts obtained from standard optical light microscope systems. Embryo selection following IVF is a critical factor in determining the success of ensuing pregnancy. Traditional morphokinetic grading by trained embryologists can be subjective and variable, and other complementary techniques, such as time-lapse imaging, require costly equipment and have not reliably demonstrated predictive ability for the endpoint of clinical pregnancy. AI methods are being investigated as a promising means for improving embryo selection and predicting implantation and pregnancy outcomes. These studies involved analysis of retrospectively collected data including standard optical light microscope images and clinical outcomes of 8886 embryos from 11 different IVF clinics, across three different countries, between 2011 and 2018. The AI-based model was trained using static two-dimensional optical light microscope images with known clinical pregnancy outcome as measured by fetal heartbeat to provide a confidence score for prediction of pregnancy. Predictive accuracy was determined by evaluating sensitivity, specificity and overall weighted accuracy, and was visualized using histograms of the distributions of predictions. Comparison to embryologists' predictive accuracy was performed using a binary classification approach and a 5-band ranking comparison. The Life Whisperer AI model showed a sensitivity of 70.1% for viable embryos while maintaining a specificity of 60.5% for non-viable embryos across three independent blind test sets from different clinics. The weighted overall accuracy in each blind test set was >63%, with a combined accuracy of 64.3% across both viable and non-viable embryos, demonstrating model robustness and generalizability beyond the result expected from chance. Distributions of predictions showed clear separation of correctly and incorrectly classified embryos. Binary comparison of viable/non-viable embryo classification demonstrated an improvement of 24.7% over embryologists' accuracy (P = 0.047, n = 2, Student's t test), and 5-band ranking comparison demonstrated an improvement of 42.0% over embryologists (P = 0.028, n = 2, Student's t test). The AI model developed here is limited to analysis of Day 5 embryos; therefore, further evaluation or modification of the model is needed to incorporate information from different time points. The endpoint described is clinical pregnancy as measured by fetal heartbeat, and this does not indicate the probability of live birth. The current investigation was performed with retrospectively collected data, and hence it will be of importance to collect data prospectively to assess real-world use of the AI model. These studies demonstrated an improved predictive ability for evaluation of embryo viability when compared with embryologists' traditional morphokinetic grading methods. The superior accuracy of the Life Whisperer AI model could lead to improved pregnancy success rates in IVF when used in a clinical setting. It could also potentially assist in standardization of embryo selection methods across multiple clinical environments, while eliminating the need for complex time-lapse imaging equipment. Finally, the cloud-based software application used to apply the Life Whisperer AI model in clinical practice makes it broadly applicable and globally scalable to IVF clinics worldwide. Life Whisperer Diagnostics, Pty Ltd is a wholly owned subsidiary of the parent company, Presagen Pty Ltd. Funding for the study was provided by Presagen with grant funding received from the South Australian Government: Research, Commercialisation and Startup Fund (RCSF). 'In kind' support and embryology expertise to guide algorithm development were provided by Ovation Fertility. J.M.M.H., D.P. and M.P. are co-owners of Life Whisperer and Presagen. Presagen has filed a provisional patent for the technology described in this manuscript (52985P pending). A.P.M. owns stock in Life Whisperer, and S.M.D., A.J., T.N. and A.P.M. are employees of Life Whisperer.

VerMilyea M ，Hall JMM ，Diakiw SM ，Johnston A ，Nguyen T ，Perugini D ，Miller A ，Picou A ，Murphy AP ，Perugini M ... -  《-》

Implicit bias in diagnosing mosaicism amongst preimplantation genetic testing providers: results from a multicenter study of 36 395 blastocysts.

Does the diagnosis of mosaicism affect ploidy rates across different providers offering preimplantation genetic testing for aneuploidies (PGT-A)? Our analysis of 36 395 blastocyst biopsies across eight genetic testing laboratories revealed that euploidy rates were significantly higher in providers reporting low rates of mosaicism. Diagnoses consistent with chromosomal mosaicism have emerged as a third category of possible embryo ploidy outcomes following PGT-A. However, in the era of mosaicism, embryo selection has become increasingly complex. Biological, technical, analytical, and clinical complexities in interpreting such results have led to substantial variability in mosaicism rates across PGT-A providers and clinics. Critically, it remains unknown whether these differences impact the number of euploid embryos available for transfer. Ultimately, this may significantly affect clinical outcomes, with important implications for PGT-A patients. In this international, multicenter cohort study, we reviewed 36 395 consecutive PGT-A results, obtained from 10 035 patients across 11 867 treatment cycles, conducted between October 2015 and October 2021. A total of 17 IVF centers, across eight PGT-A providers, five countries and three continents participated in the study. All blastocysts were tested using trophectoderm biopsy and next-generation sequencing. Both autologous and donation cycles were assessed. Cycles using preimplantation genetic testing for structural rearrangements were excluded from the analysis. The PGT-A providers were randomly categorized (A to H). Providers B, C, D, E, F, G, and H all reported mosaicism, whereas Provider A reported embryos as either euploid or aneuploid. Ploidy rates were analyzed using multilevel mixed linear regression. Analyses were adjusted for maternal age, paternal age, oocyte source, number of embryos biopsied, day of biopsy, and PGT-A provider, as appropriate. We compared associations between genetic testing providers and PGT-A outcomes, including the number of chromosomally normal (euploid) embryos determined to be suitable for transfer. The mean maternal age (±SD) across all providers was 36.2 (±5.2). Our findings reveal a strong association between PGT-A provider and the diagnosis of euploidy and mosaicism. Amongst the seven providers that reported mosaicism, the rates varied from 3.1% to 25.0%. After adjusting for confounders, we observed a significant difference in the likelihood of diagnosing mosaicism across providers (P < 0.001), ranging from 6.5% (95% CI: 5.2-7.4%) for Provider B to 35.6% (95% CI: 32.6-38.7%) for Provider E. Notably, adjusted euploidy rates were highest for providers that reported the lowest rates of mosaicism (Provider B: euploidy, 55.7% (95% CI: 54.1-57.4%), mosaicism, 6.5% (95% CI: 5.2-7.4%); Provider H: euploidy, 44.5% (95% CI: 43.6-45.4%), mosaicism, 9.9% (95% CI: 9.2-10.6%)); and Provider D: euploidy, 43.8% (95% CI: 39.2-48.4%), mosaicism, 11.0% (95% CI: 7.5-14.5%)). Moreover, the overall chance of having at least one euploid blastocyst available for transfer was significantly higher when mosaicism was not reported, when we compared Provider A to all other providers (OR = 1.30, 95% CI: 1.13-1.50). Differences in diagnosing and interpreting mosaic results across PGT-A laboratories raise further concerns regarding the accuracy and relevance of mosaicism predictions. While we confirmed equivalent clinical outcomes following the transfer of mosaic and euploid blastocysts, we found that a significant proportion of mosaic embryos are not used for IVF treatment. Due to the retrospective nature of the study, associations can be ascertained, however, causality cannot be established. Certain parameters such as blastocyst grade were not available in the dataset. Furthermore, certain platform-related and clinic-specific factors may not be readily quantifiable or explicitly captured in our dataset. As such, a full elucidation of all potential confounders accounting for variability may not be possible. Our findings highlight the strong need for standardization and quality assurance in the industry. The decision not to transfer mosaic embryos may ultimately reduce the chance of success of a PGT-A cycle by limiting the pool of available embryos. Until we can be certain that mosaic diagnoses accurately reflect biological variability, reporting mosaicism warrants utmost caution. A prudent approach is imperative, as it may determine the difference between success or failure for some patients. This work was supported by the Torres Quevedo Grant, awarded to M.P. (PTQ2019-010494) by the Spanish State Research Agency, Ministry of Science and Innovation, Spain. M.P., L.B., A.R.L., A.L.R.d.C.L., N.P.P., M.P., D.S., F.A., A.P., B.M., L.D., F.V.M., D.S., M.R., E.P.d.l.B., A.R., and R.V. have no competing interests to declare. B.L., R.M., and J.A.O. are full time employees of IB Biotech, the genetics company of the Instituto Bernabeu group, which performs preimplantation genetic testing. M.G. is a full time employee of Novagen, the genetics company of Cegyr, which performs preimplantation genetic testing. N/A.

Popovic M ，Borot L ，Lorenzon AR ，Lopes ALRC ，Sakkas D ，Lledó B ，Morales R ，Ortiz JA ，Polyzos NP ，Parriego M ，Azpiroz F ，Galain M ，Pujol A ，Menten B ，Dhaenens L ，Vanden Meerschaut F ，Stoop D ，Rodriguez M ，de la Blanca EP ，Rodríguez A ，Vassena R ... -  《-》

The use of copy number loads to designate mosaicism in blastocyst stage PGT-A cycles: fewer is better.

How well can whole chromosome copy number analysis from a single trophectoderm (TE) biopsy predict true mosaicism configurations in human blastocysts? When a single TE biopsy is tested, wide mosaicism thresholds (i.e. 20-80% of aneuploid cells) increase false positive calls compared to more stringent ones (i.e. 30-70% of aneuploid cells) without improving true detection rate, while binary classification (aneuploid/euploid) provides the highest diagnostic accuracy. Next-generation sequencing-based technologies for preimplantation genetic testing for aneuploidies (PGT-A) allow the identification of intermediate chromosome copy number alterations potentially associated with chromosomal mosaicism in TE biopsies. Most validation studies are based on models mimicking mosaicism, e.g. mixtures of cell lines, and cannot be applied to the clinical interpretation of TE biopsy specimens. The accuracy of different mosaicism diagnostic thresholds was assessed by comparing chromosome copy numbers in multiple samples from each blastocyst. Enrolled embryos were donated for research between June 2019 and September 2020. The Institutional Review Board at the Near East University approved the study (project: YDU/2019/70-849). Embryos showing euploid/aneuploid mosaicism (n = 53), uniform chromosomal alterations (single or multiple) (n = 25), or uniform euploidy (n = 39) in their clinical TE biopsy were disaggregated into five portions: the inner cell mass (ICM) and four TE segments. Collectively, 585 samples from 117 embryos were analysed. Donated blastocysts were warmed, allowed to re-expand, and disaggregated in TE portions and ICM. PGT-A analysis was performed using Ion ReproSeq PGS kit and Ion S5 sequencer (ThermoFisher). Sequencing data were blindly analysed with Ion Reporter software to estimate raw chromosome copy numbers. Intra-blastocyst comparison of copy number data was performed employing different thresholds commonly used for mosaicism detection. From copy number data, different case scenarios were created using more stringent (30-70%) or less stringent criteria (20-80%). Categorical variables were compared using the two-sample z test for proportions. When all the five biopsies from the same embryo were analysed with 30-70% thresholds, only 8.4% (n = 14/166) of patterns abnormal in the original analysis revealed a true mosaic configuration, displaying evidence of reciprocal events (3.6%, n = 6/166) or confirmation in additional biopsies (4.8%, n = 8/166), while most mosaic results (87.3% of total predicted mosaic patterns) remained confined to a single TE specimen. Conversely, uniform whole chromosome aneuploidies (28.3% of total patterns, n = 47/166) were confirmed in all subsequent biopsies in 97.9% of cases (n = 46/47). When 20-80% thresholds were employed (instead of 30-70%), the overall mosaicism rate per biopsy increased from 20.2% (n = 114/565) to 40.2% (n = 227/565). However, the use of a wider threshold range did not contribute to the detection of additional true mosaic patterns, while significantly increasing false positive mosaic patterns from 57.8% to 79.5% (n = 96/166; 95% CI = 49.9-65.4 vs n = 271/341; 95% CI = 74.8-83.6, respectively) (P < 0.00001). Moreover, the shift of the aneuploid cut-off from 70% to 80% of aneuploid cells resulted in mosaicism overcalling in the high range (50-80% of aneuploid cells), impacting the accuracy of uniform aneuploid classification. Parametric analysis of thresholds, based on multifocal analysis, revealed that a binary classification scheme with a single cut-off at a 50% level provided the highest sensitivity and specificity rates. Further analysis on technical noise distribution at the chromosome level revealed a greater impact on smaller chromosomes. While enrolment of a population enriched in embryos showing intermediate chromosome copy numbers enhanced the evaluation of the mosaicism category compared with random sampling such study population selection is likely to lead to an overall underestimation of PGT-A accuracy compared to a general assessment of unselected clinical samples. This approach involved the analysis of aneuploidy chromosome copy number thresholds at the embryo level; future studies will need to evaluate these criteria in relation to clinical predictive values following embryo transfers for different PGT-A assays. Moreover, the study lacked genotyping-based confirmation analysis. Finally, aneuploid embryos with known meiotic partial deletion/duplication were not included. Current technologies can detect low-intermediate chromosome copy numbers in preimplantation embryos but their identification is poorly correlated with consistent propagation of the anomaly throughout the embryo or with negative clinical consequences when transferred. Therefore, when a single TE biopsy is analysed, diagnosis of chromosomal mosaicism should be evaluated carefully. Indeed, the use of wider mosaicism thresholds (i.e. 20-80%) should be avoided as it reduces the overall PGT-A diagnostic accuracy by increasing the risk of false positive mosaic classification and false negative aneuploid classification. From a clinical perspective, this approach has negative consequences for patients as it leads to the potential deselection of normal embryos for transfer. Moreover, a proportion of uniform aneuploid embryos may be inaccurately categorized as high-level mosaic, with a consequent negative outcome (i.e. miscarriage) when inadvertently selected for transfer. Clinical outcomes following PGT-A are maximized when a 50% threshold is employed as it offers the most accurate diagnostic approach. The study was supported by Igenomix. The authors not employed by Igenomix have no conflicts of interest to declare. N/A.

Girardi L ，Figliuzzi M ，Poli M ，Serdarogullari M ，Patassini C ，Caroselli S ，Pergher I ，Cogo F ，Coban O ，Boynukalin FK ，Bahceci M ，Navarro R ，Rubio C ，Findikli N ，Simón C ，Capalbo A ... -  《-》

Improved clinical utility of preimplantation genetic testing through the integration of ploidy and common pathogenic microdeletions analyses.

Can chromosomal abnormalities beyond copy-number aneuploidies (i.e. ploidy level and microdeletions (MDs)) be detected using a preimplantation genetic testing (PGT) platform? The proposed integrated approach accurately assesses ploidy level and the most common pathogenic microdeletions causative of genomic disorders, expanding the clinical utility of PGT. Standard methodologies employed in preimplantation genetic testing for aneuploidy (PGT-A) identify chromosomal aneuploidies but cannot determine ploidy level nor the presence of recurrent pathogenic MDs responsible for genomic disorders. Transferring embryos carrying these abnormalities can result in miscarriage, molar pregnancy, and intellectual disabilities and developmental delay in offspring. The development of a testing strategy that integrates their assessment can resolve current limitations and add valuable information regarding the genetic constitution of embryos, which is not evaluated in PGT providing new level of clinical utility and valuable knowledge for further understanding of the genomic causes of implantation failure and early pregnancy loss. To the best of our knowledge, MDs have never been studied in preimplantation human embryos up to date. This is a retrospective cohort analysis including blastocyst biopsies collected between February 2018 and November 2021 at multiple collaborating IVF clinics from prospective parents of European ancestry below the age of 45, using autologous gametes and undergoing ICSI for all oocytes. Ploidy level determination was validated using 164 embryonic samples of known ploidy status (147 diploids, 9 triploids, and 8 haploids). Detection of nine common MD syndromes (-4p=Wolf-Hirschhorn, -8q=Langer-Giedion, -1p=1p36 deletion, -22q=DiGeorge, -5p=Cri-du-Chat, -15q=Prader-Willi/Angelman, -11q=Jacobsen, -17p=Smith-Magenis) was developed and tested using 28 positive controls and 97 negative controls. Later, the methodology was blindly applied in the analysis of: (i) 100 two pronuclei (2PN)-derived blastocysts that were previously defined as uniformly euploid by standard PGT-A; (ii) 99 euploid embryos whose transfer resulted in pregnancy loss. The methodology is based on targeted next-generation sequencing of selected polymorphisms across the genome and enriched within critical regions of included MD syndromes. Sequencing data (i.e. allelic frequencies) were analyzed by a probabilistic model which estimated the likelihood of ploidy level and MD presence, accounting for both sequencing noise and population genetics patterns (i.e. linkage disequilibrium, LD, correlations) observed in 2504 whole-genome sequencing data from the 1000 Genome Project database. Analysis of phased parental haplotypes obtained by single-nucleotide polymorphism (SNP)-array genotyping was performed to confirm the presence of MD. In the analytical validation phase, this strategy showed extremely high accuracy both in ploidy classification (100%, CI: 98.1-100%) and in the identification of six out of eight MDs (99.2%, CI: 98.5-99.8%). To improve MD detection based on loss of heterozygosity (LOH), common haploblocks were analyzed based on haplotype frequency and LOH occurrence in a reference population, thus developing two further mathematical models. As a result, chr1p36 and chr4p16.3 regions were excluded from MD identification due to their poor reliability, whilst a clinical workflow which incorporated parental DNA information was developed to enhance the identification of MDs. During the clinical application phase, one case of triploidy was detected among 2PN-derived blastocysts (i) and one pathogenic MD (-22q11.21) was retrospectively identified among the biopsy specimens of transferred embryos that resulted in miscarriage (ii). For the latter case, family-based analysis revealed the same MD in different sibling embryos (n = 2/5) from non-carrier parents, suggesting the presence of germline mosaicism in the female partner. When embryos are selected for transfer based on their genetic constitution, this strategy can identify embryos with ploidy abnormalities and/or MDs beyond aneuploidies, with an estimated incidence of 1.5% (n = 3/202, 95% CI: 0.5-4.5%) among euploid embryos. Epidemiological studies will be required to accurately assess the incidence of ploidy alterations and MDs in preimplantation embryos and particularly in euploid miscarriages. Despite the high accuracy of the assay developed, the use of parental DNA to support diagnostic calling can further increase the precision of the assay. This novel assay significantly expands the clinical utility of PGT-A by integrating the most common pathogenic MDs (both de novo and inherited ones) responsible for genomic disorders, which are usually evaluated at a later stage through invasive prenatal testing. From a basic research standpoint, this approach will help to elucidate fundamental biological and clinical questions related to the genetics of implantation failure and pregnancy loss of otherwise euploid embryos. No external funding was used for this study. S.C., M.F., F.C., P.Z., I.P., L.G., C.P., M.P., D.B., J.J.-A., D.B.-J., J.M.-V., and C.R. are employees of Igenomix and C.S. is the head of the scientific board of Igenomix. A.C. and L.P. are employees of JUNO GENETICS. Igenomix and JUNO GENETICS are companies providing reproductive genetic services. N/A.

Caroselli S ，Figliuzzi M ，Picchetta L ，Cogo F ，Zambon P ，Pergher I ，Girardi L ，Patassini C ，Poli M ，Bakalova D ，Cimadomo D ，Findikli N ，Coban O ，Serdarogullari M ，Favero F ，Bortolato S ，Anastasi A ，Capodanno F ，Gallinelli A ，Brancati F ，Rienzi L ，Ubaldi FM ，Jimenez-Almazán J ，Blesa-Jarque D ，Miravet-Valenciano J ，Rubio C ，Simòn C ，Capalbo A ... -  《-》