New AI-based system significantly reduces IVF-related miscarriages, genetic disorders

Since its inception in 1978, in vitro fertilization (IVF) has provided a pathway to parenthood for millions, with over 8 million successful births attributed to this groundbreaking technology. At its core, IVF hinges on selecting the best-quality embryos, a task historically reliant on manual assessments.

Advances in artificial intelligence (AI), such as the new system called BELA (Blastocyst Evaluation Learning Algorithm), promise to revolutionize this process by providing automated, accurate predictions of embryo quality and chromosomal health, known as ploidy status.

Ploidy status is a critical determinant of an embryo’s viability. Euploid embryos, with the correct number of chromosomes, are more likely to lead to successful pregnancies.

Conversely, aneuploid embryos, marked by chromosomal abnormalities, are often associated with miscarriage, implantation failure, or genetic disorders such as Down syndrome. The risk of aneuploidy increases with maternal age, posing significant challenges for older individuals undergoing IVF.

Traditionally, preimplantation genetic testing for aneuploidy (PGT-A) has been used to assess ploidy. While effective, PGT-A requires invasive procedures to biopsy embryo cells, potentially compromising viability.

It is also expensive, time-consuming, and susceptible to inaccuracies due to embryonic mosaicism, where an embryo contains both euploid and aneuploid cells. These limitations have driven researchers to explore non-invasive alternatives.

The advent of AI and the accumulation of extensive IVF datasets have opened the door to automated embryo assessment. By analyzing time-lapse images of embryos, AI models can identify subtle patterns that may indicate ploidy status, eliminating the need for invasive testing. BELA represents a significant step forward in this field.

Developed by researchers at Weill Cornell Medicine and published in the journal, Nature Communications, BELA is a machine-learning model that leverages time-lapse video sequences of embryo development and maternal age to predict ploidy status. Unlike earlier models, BELA eliminates the need for embryologists' subjective assessments, providing an objective and scalable solution.

BELA operates in two phases. First, it predicts the blastocyst score (BS), a measure of embryo quality, by analyzing time-lapse videos from the critical period between 96 and 112 hours post-insemination (hpi). The model’s predictions align closely with those of experienced embryologists, with a mean absolute error of 1.855.

This accuracy reflects the model’s ability to identify key developmental milestones, such as the timing of blastocyst expansion, which are crucial for assessing embryo quality.

In the second phase, BELA combines the model-derived blastocyst score (MDBS) with maternal age to predict ploidy status. The model distinguishes between euploid and aneuploid embryos with an area under the receiver operating characteristic curve (AUC) of 0.76, outperforming earlier AI systems. By integrating maternal age, BELA improves its predictive accuracy, reflecting clinical practices where age is a significant factor in embryo assessment.

BELA’s design incorporates advanced techniques like BiLSTM (bidirectional long short-term memory) modeling to capture the sequential nature of embryo development. This approach enables the system to assess changes over time, mirroring the methodology used by skilled embryologists. By focusing on specific developmental stages, BELA minimizes noise from irrelevant data, enhancing its predictive power.

To validate BELA, researchers used datasets from Weill Cornell Medicine’s Embryoscope® and Embryoscope+® systems, encompassing nearly 2,000 embryos. The model was further tested on external datasets from clinics in Florida and Spain.

Across all datasets, BELA demonstrated robust performance, surpassing previous models in accuracy and generalizability. For instance, when distinguishing between euploid and complex aneuploid embryos, BELA achieved an AUC of 0.826 with maternal age factored in.

Notably, BELA’s reliance on time-lapse sequences rather than static images enhances its ability to capture dynamic developmental patterns. This approach mirrors the methods used by embryologists, who assess the speed and timing of blastulation to infer embryo quality. By automating this process, BELA offers a consistent and objective alternative to manual evaluation.

The model’s integration of maternal age as a feature further strengthens its clinical relevance. Maternal age is a well-documented factor influencing embryo quality, and its inclusion in BELA’s algorithms aligns with real-world practices. Supplementary analyses revealed that lower maternal age correlates with higher euploid predictions, reflecting established biological principles.

BELA’s development marks a significant milestone in IVF technology. By eliminating the need for invasive testing, it could reduce the costs and risks associated with PGT-A, making IVF more accessible to a broader population.

“BELA and AI models like it could expand the availability of IVF to areas that don’t have access to high-end IVF technology and PGT testing, improving equity in IVF care across the world,” said Dr. Nikica Zaninovic, director of the Embryology Laboratory at Weill Cornell Medicine.

The system’s ability to process vast amounts of image data also suggests broader applications. Beyond ploidy prediction, BELA could be adapted for general embryo quality assessment or to predict developmental stages. As Dr. Suraj Rajendran, a doctoral researcher involved in the project, noted, “Our hope is that this model could be useful for various functions that an embryology clinic could tailor to its needs.”

In addition to expanding access to IVF, BELA has the potential to improve success rates by streamlining embryo selection. Traditional methods rely heavily on subjective assessments, which can vary between clinics and practitioners. By providing a standardized, data-driven approach, BELA ensures consistency in evaluating embryo quality, reducing the likelihood of errors.

Despite its promise, BELA is not a replacement for PGT-A but a complementary tool that could streamline the embryo evaluation process. The researchers are planning clinical trials to test BELA’s predictive power in real-world settings. If successful, these trials could pave the way for widespread adoption, transforming how IVF clinics approach embryo selection.

BELA’s development also underscores the potential of AI in reproductive medicine. By combining advanced algorithms with clinical expertise, researchers are pushing the boundaries of what is possible in fertility care. As Dr. Iman Hajirasouliha, a senior researcher on the project, explained, “This is a fully automated and more objective approach compared to prior methods, offering greater predictive power through the use of extensive image data.”

The potential for BELA extends beyond embryo ploidy prediction. Researchers envision its application in other areas of embryology, such as predicting implantation success or identifying embryos at risk of developmental issues. By leveraging AI’s ability to analyze complex datasets, BELA could revolutionize multiple aspects of fertility treatment.

The integration of AI into IVF represents a paradigm shift in fertility care. With systems like BELA, clinics can achieve greater efficiency, accuracy, and accessibility, ultimately improving outcomes for patients worldwide.

While challenges remain, the progress made by researchers at Weill Cornell Medicine highlights the transformative potential of technology in addressing one of the most personal and profound challenges faced by individuals and families.

By reducing costs, minimizing risks, and enhancing predictive accuracy, BELA exemplifies the promise of AI-driven solutions in healthcare. As clinical trials progress, the future of IVF looks increasingly bright, offering hope to countless individuals striving to build families.

Note: Materials provided above by The Brighter Side of News. Content may be edited for style and length.

Like these kind of feel good stories? Get The Brighter Side of News' newsletter.