NEWS
Backed by proven science
This study explored whether AI can predict fetal heartbeat in frozen-thawed embryos using post-warming time-lapse images and videos. A positive correlation was found between AI scores and fetal heartbeat outcomes, with the best results in non-genetically tested embryos. Prediction accuracy reached up to 74% in optimal conditions. AI outperformed traditional morphology assessments and identified embryos […]
This study validated the EMBRYOAID app, which uses AI to score embryos from photos or videos. The scores correlated with embryo morphology, development speed, euploidy, and implantation outcomes, especially in treatments using patient or donor oocytes. Higher scores were linked to better morphology, faster development, and increased chances of implantation. The tool performed well in […]
This international, randomized, multicenter study compares AI-based embryo assessment using the EMBRYOAID tool with standard evaluation by experienced embryologists. Embryos in the control group were selected based on the Gardner scale, while those in the test group were chosen according to AI recommendations. Results show that AI-supported embryo selection can achieve pregnancy rates comparable to […]
This study evaluated the accuracy of automated ovarian follicle measurements using the FOLLISCAN AI platform integrated into routine IVF practice. A total of 294 ultrasound videos from 147 exams involving 101 patients were analyzed, resulting in 4,347 follicle annotations. The findings suggest that AI-based follicle annotation offers consistent and efficient measurements, requiring minimal expert intervention […]
This study explored the use of AI to support trigger day decisions in IVF, aiming to improve the number of mature oocytes retrieved. An algorithm analyzed ultrasound and clinical data to recommend the optimal day for triggering follicular maturation. The study compared outcomes between cases where AI recommendations aligned with physician decisions and those where […]
This study examined whether automated follicle measurements using AI can predict the number of retrieved and mature oocytes as accurately as traditional manual assessments. Data from IVF cycles across five centers were analyzed, comparing physician-reported measurements with those generated by an AI platform. Both types of data were used in predictive models to estimate oocyte […]
This study evaluated whether an AI model can match the performance of experienced embryologists in selecting embryos with the highest likelihood of implantation. Using a test of 150 embryo pairs, where one embryo in each pair led to pregnancy, both the AI and a group of expert embryologists were asked to choose the more promising […]
This study addressed the instability of AI models used in embryo selection, which can be sensitive to small, irrelevant changes in input images. A deep learning model was tested on various image modifications, such as rotation and brightness shifts, and applied techniques like ensembling and robust training. These methods reduced score variability by 86% without […]
This study evaluated whether a simple, interpretable feature, the blastocyst area, can match the performance of advanced AI models in embryo ranking. Using a large, standardized dataset, it was found that ranking embryos by size alone produced results comparable to state-of-the-art AI methods. Given its objectivity and ease of use, embryo area could serve as […]
This study explored the use of deep learning to automate follicle detection and measurement in ultrasound cine-loops. A 3D neural network model was trained on clinical data and performed well in identifying and outlining follicles without manual input. Its accuracy was close to that of human experts, particularly for larger follicles during key stimulation days. […]
This study developed a model that combines clinical and genetic data to predict the number of MII oocytes retrieved during ovarian stimulation. Using advanced machine learning on large clinical and smaller clinical-genetic datasets, the combined model outperformed one based on clinical data alone. Key predictors included hormone levels, follicle count, and genetic variants in several […]
This paper compares two game-theoretic methods, Shapley and Banzhaf values, or explaining tree ensemble model predictions. It introduces a faster algorithm for computing Banzhaf values, which run more efficiently than the current Shapley-based methods. Both methods provide similar feature importance scores, but the Banzhaf value is not only quicker to compute but also more numerically […]
This study explored whether genetic factors influence the number of MII oocytes retrieved during ovarian stimulation. Using data from 516 stimulations and analyzing gene variants, specific haplotypes were identified that improved prediction accuracy beyond clinical factors alone. A neural network model including clinical data served as a benchmark, and adding key genetic variants reduced prediction […]
This content is intended exclusively for qualified medical professionals and may include specialized information that should not be interpreted by individuals without appropriate medical training.
