The need to reduce the number of embryos transferred in assisted reproductive care to prevent multiple gestations has led to a stronger emphasis on selecting embryos with the highest morphological quality. Although this evaluation has traditionally been performed by trained embryologists, the increasing use of time-lapse incubators has introduced a greater volume of data and subjectivity in decision-making. Artificial intelligence (AI)-based tools can support embryologists by offering objective, standardized embryo assessments.In Brazil, like other countries, where imported embryo selection technologies may not account for local demographic and ethnic profiles, an AI model — Morphological Artificial Intelligence Assistance (MAIA) — was developed through a collaboration between a university and a private fertility clinic in São Paulo. The model was trained using 1,015 embryo images and prospectively tested in a clinical setting on 200 single embryo transfers. In clinical testing, MAIA achieved an overall accuracy of 66.5%. In elective embryo transfers, where there were more than one embryo eligible for transfer, MAIA achieved 70.1% accuracy for predicting clinical pregnancy. Designed with a user-friendly interface tailored by embryologists, MAIA provides real-time embryo evaluations to support decision-making in routine care.

Background: In recent times, various algorithms have been developed to assist in the selection of embryos fortransfer based on artificial intelligence (AI). Nevertheless, the majority of AI models employed in this context werecharacterized by a lack of transparency. To address these concerns, we aim to design an interpretable tool to automatehuman embryo evaluation by combining artificial neural networks (ANNs) and genetic algorithms (GA).Materials and Methods: This retrospective cohort study included 223 human blastocyst time-lapse (TL) imagestaken at 110 hours post-injection. All the images were evaluated by five embryologists from different clinics in termsof blastocyst expansion (BE), quality of the inner cell mass (ICM), and trophectoderm (TE). The embryo databasewas used to develop an AI system (70% training, 15% validation, and 15% test) for automate blastocyst assessment.The entire set of images underwent a standardization process, followed by processing and segmentation using Matlabsoftware. The resulting quantified variables were utilized in AI techniques (ANN and GA). Finally, the accuracy andperformance of the automation tool was assessed with the area under the receiver operating characteristic (ROC)curve (AUC). Then, the level of agreement among embryologists and between embryologists and the AI system wascompared with Kappa Index.Results: The overall agreement among embryologists was low (Kappa: 0.4 for BE; and 0.3 for TE and ICM). The AItool achieved higher consistency (Kappa 0.7 for BE and ICM; and 0.4 for TE). The AI exhibited high accuracy in classifyingBE (test 81.5%), ICM (test 78.8%), and TE (test 78.3%) and better performance for BE (AUC 0.888-0.956)than for ICM (AUC 0.605-0.854) and TE (AUC 0.726-0.769) assessment.Conclusion: Our AI tool highlighted the superior consistency of AI compared to human operators in grading blastocystmorphology. This research represents an important step towards fully automating objective embryo evaluation.

Thoroughly analyzing the sperm and exploring the information obtained using artificial intelligence (AI) could be the key to improving fertility estimation. Artificial neural networks have already been applied to calculate zootechnical indices in animals and predict fertility in humans. This method of estimating the results of reproductive biotechnologies, such as in vitro embryo production (IVEP) in cattle, could be valuable for livestock production. This study was developed to model IVEP estimates in Senepol animals based on various sperm attributes, through retrospective data from 290 IVEP routines performed using 38 commercial doses of semen from Senepol bulls. All sperm samples that had undergone the same procedure during sperm selection for in vitro fertilization were evaluated using a computer-assisted sperm analysis (CASA) system to define sperm subpopulations. Sperm morphology was also analyzed in a wet preparation, and the integrity of the plasma and acrosomal membranes, mitochondrial potential, oxidative status, and chromatin resistance were evaluated using flow cytometry. A previous study identified three sperm subpopulations in such samples and the information used in tandem with other sperm quality variables to perform an AI analysis. AI analysis generated models that estimated IVEP based on the season, donor, percentage of viable oocytes, and 18 other sperm predictor variables. The accuracy of the results obtained for the three best AI models for predicting the IVEP was 90.7, 75.3, and 79.6%, respectively. Therefore, applying this AI technique would enable the estimation of high or low embryo production for individual bulls based on the sperm analysis information.

To predict the ability of a vitrified/warmed blastocyst to lead to a live birth by applying artificial intelligence on time-lapse images recorded from devitrification to embryo transfer.

Despite the use of new techniques on embryo selection and the presence of equipment on the market, such as EmbryoScope® and Geri®, which help in the evaluation of embryo quality, there is still a subjectivity between the embryologist’s classifications, which are subjected to inter- and intra-observer variability, therefore compromising the successful implantation of the embryo. Nonetheless, with the acquisition of images through the time-lapse system, it is possible to perform digital processing of these images, providing a better analysis of the embryo, in addition to enabling the automatic analysis of a large volume of information. An image processing protocol was developed using well-established techniques to segment the image of blastocysts and extract variables of interest. A total of 33 variables were automatically generated by digital image processing, each one representing a different aspect of the embryo and describing a different characteristic of the blastocyst. These variables can be categorized into texture, gray-level average, gray-level standard deviation, modal value, relations, and light level. The automated and directed steps of the proposed processing protocol exclude spurious results, except when image quality (e.g., focus) prevents correct segmentation. The image processing protocol can segment human blastocyst images and automatically extract 33 variables that describe quantitative aspects of the blastocyst’s regions, with potential utility in embryo selection for assisted reproductive technology (ART).

Morphology, morphokinetics, preimplantation genetic screening and lately artificial intelligence (AI) algorithms are used to rank and select embryos with higher potential to achieve a live birth. Training and validating an AI for embryo selection evolves the analysis of their pregnancy prediction potential out of a transferred embryo. In order to truly understand the efficiency of an in development algorithm , we compared senior (>8 years of experience) embryologists (EMBs) and an AI algorithm abilities for pregnancy prediction in a cohort of single-embryos transferred.

The leading cause of implantation failure from embryonic origin in humans is aneuploidy. Embryo biopsy and preimplantation genetic testing is the current standard technique to assess embryo ploidy, however, it is an invasive, likely harmful and expensive method. The aim of this study is to develop an alternative, non-invasive method, to predict blastocyst ploidy using an artificial neural network (ANN) algorithm built and trained with biopsied embryo for PGT-A images and patient's parameters. Prospective cohort study including patients undergoing in vitro fertilization (IVF) treatment and blastocyst biopsy (NGS platform) after inform consent form signature (n=118 patients, n=408 biopsied blastocysts). All embryos were cultured in a time-lapse incubator (Embryoscope Plus, Vitrolife) between July 2019 and September 2020. Blastocyst digital image processing were analyzed considering 33 mathematic variables. Patients parameters considered for each embryo were: maternal age, body mass index, number of oocytes retrieved, number of previous cycles, percentage of fragmented cells at 2-cell stage, number of multi/binucleated cells at 2-cell stage, oocyte source (fresh or frozen) and the presence of specific infertility factors (endometriosis, male factor and tubal factor). Input data were randomized for training, validation and test (70, 15 and 15%, respectively). Three ANN algorithms were selected (with the aid of a genetic algorithm), trained and validated, according to input data type: ANN1: considered blastocyst image only (morphology); ANN2: considered patient parameters only and ANN3: considered both morphology and patient parameters. The area under the curve (AUC) of the receiver operating characteristic curve was measured to obtain predictive power. ANN1 was trained and validated with 148 embryos (AUC for both euploid and aneuploid= 0.99) and tested with 42 embryos (AUC for euploid=0.74 and aneuploidy=0.62). ANN2 was trained and validated with 125 embryos (AUC for euploid=0.90 and aneuploidy=0.89) and tested with 23 embryos (AUC for euploid=0.72 and aneuploidy=0.78). Considering both the morphology and patient parameters (ANN3), for training and validation were used data from 100 embryos (AUC for euploid=0.97 and aneuploidy=0.96) and 18 for testing (AUC for euploid=0.83 and aneuploidy=0.85). Sixty-two embryos were excluded from dataset due their poor image quality. The multidisciplinary approach has been employed to develop new technologies for assisted reproductive medicine. The use of ANN for embryo assessment is a promising tool to be used in IVF laboratories. In our model, using both morphology and patient parameters, it was achieved a higher predictive power to evaluate euploid and aneuploid embryos, which may be potentially used to select the best embryo to achieve a clinical pregnancy.

Research question: The study aimed to develop an artificial intelligence model based on artificial neural networks (ANNs) to predict the likelihood of achieving a live birth using the proteomic profile of spent culture media and blastocyst morphology. Design: This retrospective cohort study included 212 patients who underwent single blastocyst transfer at IVI Valencia. A single image of each of 186 embryos was studied, and the protein profile was analysed in 81 samples of spent embryo culture medium from patients included in the preimplantation genetic testing programme. The information extracted from the analyses was used as input data for the ANN. The multilayer perceptron and the back-propagation learning method were used to train the ANN. Finally, predictive power was measured using the area under the curve (AUC) of the receiver operating characteristic curve. Results: Three ANN architectures classified most of the embryos correctly as leading (LB+) or not leading (LB-) to a live birth: 100.0% for ANN1 (morphological variables and two proteins), 85.7% for ANN2 (morphological variables and seven proteins), and 83.3% for ANN3 (morphological variables and 25 proteins). The artificial intelligence model using information extracted from blastocyst image analysis and concentrations of interleukin-6 and matrix metalloproteinase-1 was able to predict live birth with an AUC of 1.0. Conclusions: The model proposed in this preliminary report may provide a promising tool to select the embryo most likely to lead to a live birth in a euploid cohort. The accuracy of prediction demonstrated by this software may improve the efficacy of an assisted reproduction treatment by reducing the number of transfers per patient. Prospective studies are, however, needed.

To introduce the analysis of blastocysts images from alternative time-lapse monitoring systems (TLM) by using Artificial Intelligence in the prediction of live birth (LB). Retrospective cohort Study. We used image analysis technology as a tool to evaluate TLM images of 244 blastocyst stage embryos at 111.5 ± 1.5h post ICSI. The embryos were cultured in Geri-TLM incubator (Genea, Australia) until day 5 of development. Of the 244 blastocysts analyzed, from a single embryo transfer program, 200 were used for training (82%) and 44 for a blind test (18%) for prediction of LB by using AI. It was built a model of artificial neuronal network (ANN) to produce a predictable outcome of LB. Several independent numerical variables extracted from standardized TLM images as an input data were used. The efficacy of prediction of live birth was quantified and assessed using confusion matrices (True Positive-TP, True Negative-TN, False Positive-FP, False Negative-FN; Positive Prediction Value-PPV and Negative Prediction Value-NPV), ROC curves and AUC. The accuracy (ACC) of prediction of live birth by AI using an ANN model was 85.2% (208/244; TP= 68, TN= 140, FP= 19, FN= 17). In the training dataset the ACC was 90.5% (181/200; TP= 58, TN= 123, FP= 14, FN= 5, AUC= 0.868), and in the blind test dataset, ACC was 61.4% (27/44; TP= 10, TN= 17, FP= 15, FN= 12, AUC= 0.634). The PPV, precision with which the ANN model was able to classify correctly a positive LB was 78.2%. In training dataset, the precision was 80.6% for LB+, and in blind test dataset 66.6%. Likewise, the overall NPV, capacity to classify correctly a negative LB, was 89.2%. In training dataset, the NPV was 96.0%, and in blind test dataset 58.6% for LB-. The AUC in the Blind test for positive and negative LB were very similar, 0.634 and 0.618 respectively. This is the first ANN model using images from Geri TLM incubator as a target in image analysis technology. The model shows a competitive accuracy, predictive power and precision to improve the efficacy of embryo selection performed by the standard morphology and increasing the odds of LB per embryo transfer.

The aim of this study is to develop an artificial intelligence (AI) model able to predict the life-birth probability of in-vitro cultured embryos, combining morphology and morphokinetic (MK) information of their development with their oxidative stress level. This retrospective study includes 131 transferred embryos (fresh and frozen-thawed) cultured individually in an Embryoscope (ESD) incubator (Vitrolife, Denmark) until day 5/6. They belong to 100 ICSI cycles performed between May 2017 and December 2018, using autologous or donated eggs. After transfer, the oxidative status of the culture media was assessed by the Thermochemioluminiscence (TCL) AnalyzerTM (Carmel diagnostics, Israel) as an indirect measurement of their metabolic activity. A machine learning algorithm was trained using MK data obtained using the time-lapse monitoring system, the TCL values and the clinical outcome of each embryo, generating a predictive model of life-birth probability. Fertilization was performed following the standard protocol of the clinic. Scoring and selection for transfer/freezing were performed according to the ASEBIR criteria, combining morphological and MK assessment. The oxidative status of 15 μL aliquots of media was measured by the TCL Analyzer, which counts the photons emitted per second (cps) as result of the heat-induced oxidation and modification of the sample. TCL parameters used were H1, H2 and H3 (TCL amplitude at 55, 155 and 255 seconds after heating, respectively), as well as their "sm" variants, resulting from applying a smoothing algorithm to normalize the data. The AI analysis was performed using machine learning algorithms, combining different sets of variables. Data from 105 embryos was used for the training of the machine, and 26 for the simulation. Properly timed development and higher TCL parameters directly related to higher chances of achieving life-birth, as previously published. Five datasets were formed using different combinations of MK and TCL parameters with the stronger statistic correlation with life-birth result: MK+H1+H2+H3, MK+H1sm+H2sm+H3sm, MK+H1sm, MK+H2sm and MK+H3sm. They were compared to a reference model using only MK. The combination with the highest predictive capacity was MK + H2sm. In the training, the model achieved a 93.6% accuracy for predicting positive life-birth (LB+) and 86.2% for negative (LB-), compared to the 97.9% LB+ and 94.8% LB- using solely MK. The blind test of MK + H2sm model had a stronger predictive power (83.3% LB+, 85.7% LB-) than only MK (66.7% LB+, 85.7% LB-). Other combinations of oxidation variables did not show an improved predictive capacity when compared to the MK algorithm. The presented AI algorithm combining MK and TCL parameter H2sm scores a higher predictive power for life-birth than using only MK data. This supports the relevance of the oxidative status of the culture media as an indirect measurement of metabolic activity, higher in embryos that achieve birth. This AI algorithm provides an upgraded score system to assist embryo selection, potentially improving clinical results.

Based on growing demand for assisted reproduction technology, improved predictive models are required to optimize in vitro fertilization/intracytoplasmatic sperm injection strategies, prioritizing single embryo transfer. There are still several obstacles to overcome for the purpose of improving assisted reproductive success, such as intra- and inter-observer subjectivity in embryonic selection, high occurrence of multiple pregnancies, maternal and neonatal complications. Here, we compare studies that used several variables that impact the success of assisted reproduction, such as blastocyst morphology and morphokinetic aspects of embryo development as well as characteristics of the patients submitted to assisted reproduction, in order to predict embryo quality, implantation or live birth. Thereby, we emphasize the proposal of an artificial intelligence-based platform for a more objective method to predict live birth.

The beer quality can be modulated from changes in their ingredient proportions, as well as in operating parameters. The crossed experimental designs and the multiple optimizations based on desirability functions have demonstrated to be effective methodologies in the unit operation polynomial modeling and optimization of bioprocess, respectively. However, artificial intelligence techniques have been used as an alternative to this modeling in bioprocess. Therefore, this study aimed to implement a software combining artificial neural network (ANN) and differential evolution to optimize the topology of an ANN to model the Ale beer production and to use the optimized ANN in ingredients and operation parameters choice that ensure a beer with high acceptance rate, by the genetic algorithm technique for multiple-objective function. This approach allowed to find ANN models which fitted the process with correlation coefficients higher than 0.85 and high satisfaction level of beer desirable quality attributes (global desirability value = 0.78). Practical Applications This manuscript could be useful for bioprocess professionals involved in the development of the brewing process and artificial intelligence applications. The approach applied in this work allows for modeling and optimization of brewing process using a combination of crossed experimental design, artificial neural networks, and evolutionary algorithms with relatively low experimental efforts. At the same time, the quality attributes of the beer are better controlled.

We develop an online graphical and intuitive interface connected to a server aiming to facilitate access to professionals worldwide that face problems with bovine blastocysts classification. The interface Blasto3Q (3Q is referred to the three qualities of the blastocyst grading) contains a description of 24 variables that are extracted from the image of the blastocyst and analyzed by three Artificial Neural Networks (ANNs) that classifies the same loaded image. The same embryo ( i.e. , the biological specimen) was submitted to digital image capture by the control group (inverted microscope with 40x of magnification) and to experimental group (stereomicroscope with maximum of magnification plus 4x zoom from the cell phone). The 36 images obtained from control and experimental groups were uploaded on the Blasto3Q. Each image from both sources was evaluated for segmentation and submitted (only if it could be properly or partially segmented) to the quality grade classification by the three ANNs of the Blasto3Q program. In the group control, all the images were properly segmented, whereas 38.9% (07/18) and 61.1% (11/18) of the images from the experimental group, respectively could not be segmented or were partially segmented. The percentage of agreement was calculated when the same blastocyst was evaluated by the same ANN from the two sources (control and experimental groups). On the 54 potential evaluations of the three ANNs ( i.e. , 18 images been evaluated by the three networks) from the experimental group only 22.2% agreed with evaluations of the control (12/54). Of the remaining 42 disagreed evaluations from experimental group, 21 were unable to be performed and 21 were wrongly processed when compared with control evaluation.

To apply Artificial Intelligence (AI) technology on time-lapse (TLM) embryo images and morphokinetic parameters to predict live birth. The morphokinetic parameters (n=131, ICSI only), with known live birth data from single blastocyst transfers, and 131 TLM images of embryos at 111.5 hours post ICSI were used to train (70%), validate (15%), and blindly test (15%) for prediction of live birth by an AI feature-extraction system. Inclusion criteria involved recipients from our oocyte donation program with single blastocyst transfer and non-PGT. Absolute and interim cleavage time points (t2 to t8) were used, along with 33 independent numerical variables extracted from standardized TLM images as an input data. The artificial neural network (ANN) architecture associated with the genetic algorithm was used to produce a predictable output of live birth. The efficacy of prediction of live birth was quantified and assessed using ROC curves, AUC and confusion matrices (True Positive -TP, True Negative -TN, False Positive -FP, and False Negative-FN). Overall accuracy of prediction of live birth by AI using morphokinetic data was 96.2% (126/131; TP= 37, TN= 69, FP= 1, FN= 4, AUC= 0.946). In the training dataset, the accuracy was 95.5% (86/91, AUC 0.96), and in the blind test dataset, accuracy was 100% (20/20, AUC=0,961). The overall accuracy of live birth by AI using image analysis was 90.1% (100/111, TP=39, TN= 61, FP= 7, FN= 4, AUC= 0.91). In the training dataset, the accuracy was 89% (81/91, AUC 0.887), and in the blind test dataset, accuracy was 95% (19/20, AUC=0.67-0.94). The combination of morphology and morphokinetics, the AUC for positive were similar (0.96) but for negative live birth were less predective (0.65). This is the first time that AI is used to evaluate human embryo quality using morphokinetic and morphological assessment in a data set of single embryo transfers from an oocyte donation program with known live birth. Our data suggests that AI can be used to enhance the efficacy of embryo selection performed by the standard morphology or the existing algorithms of morphokinetics. Applying AI in conjunction with morphokinetic or image analysis has the potential for being the platform of embryo selection, with similar predictive abilities when treated independently although its combination may not improve the performance of AI.

To determine whether convolutional neural network (CNN) can be used to predict whether an embryo capable of achieving a pregnancy will ultimately miscarry or lead to live birth based on Artificial Intelligence (AI) analysis of time-lapse (TLM) embryo images.

The water quality index (WQI) is an important tool for water resource management and planning. However, it has major disadvantages: the generation of chemical waste, is costly, and time-consuming. In order to overcome these drawbacks, we propose to simplify this index determination by replacing traditional analytical methods with ultraviolet-visible (UV–Vis) spectrophotometry associated with artificial neural network (ANN). A total of 100 water samples were collected from two rivers located in Assis, SP, Brazil and calculated the WQI by the conventional method. UV–Vis spectral analyses between 190 and 800 nm were also performed for each sample followed by principal component analysis (PCA) aiming to reduce the number of variables. The scores of the principal components were used as input to calibrate a three-layer feed-forward neural network. Output layer was defined by the WQI values. The modeling efforts showed that the optimal ANN architecture was 19-16-1, trainlm as training function, root–mean–square error (RMSE) 0.5813, determination coefficient between observed and predicted values ( R 2 ) of 0.9857 ( p < 0.0001), and mean absolute percentage error (MAPE) of 0.57% ± 0.51%. The implications of this work’s results open up the possibility to use a portable UV–Vis spectrophotometer connected to a computer to predict the WQI in places where there is no required infrastructure to determine the WQI by the conventional method as well as to monitor water body’s in real time.