AI algorithms for non-invasive screening of IVF embryos
Release date:
2023-09-20
Beijing, December 25 (Science and Technology Daily) — According to a study published in the latest issue of The Lancet Digital Health, researchers at Weill Cornell Medicine have developed a new artificial intelligence (AI) algorithm that overcomes the limitations of biopsy by noninvasively determining whether the chromosome count of in vitro‑fertilized embryos is normal, with an accuracy rate of approximately 70 percent.
BEIJING, Dec. 25 (Science and Technology Daily) — According to a study published in the latest issue of The Lancet Digital Health, researchers at Weill Cornell Medicine have developed a new artificial intelligence (AI) algorithm that overcomes the limitations of biopsy by noninvasively determining whether the chromosome count of in vitro‑fertilized embryos is normal, with an accuracy rate of approximately 70%.
Chromosome number abnormalities, known as aneuploidy, are a leading cause of embryo implantation failure or the inability to achieve a healthy pregnancy in in vitro fertilization (IVF). One current method for detecting aneuploidy involves taking a biopsy‑like sample from embryonic cells and performing genetic analysis; this approach increases the cost of IVF and is invasive to the embryo.
Currently, physicians primarily use microscopy to assess whether embryos exhibit significant abnormalities associated with poor viability. To obtain chromosomal information, clinicians may also employ a biopsy-based technique known as preimplantation aneuploidy testing (PGT-A).
In a new study, the research team developed the STORK-A algorithm as a potential alternative to PGT‑A, or as a more selective approach for determining which embryos should undergo PGT‑A testing.
The new algorithm, STORK‑A, utilizes embryo microscopy images captured on day 5 post‑fertilization, embryo quality scores, maternal age, and other relevant information to automatically “learn” how to associate specific features of the data with the likelihood of aneuploidy. The research team trained STORK‑A on a dataset of 10,378 blastocysts whose ploidy status was already known. They then tested the algorithm on an independent dataset, achieving comparable accuracy and demonstrating the generalizability of STORK‑A.
According to researchers’ assessments, the algorithm achieves nearly 70% accuracy in predicting aneuploid embryos versus euploid, or “normal‑chromosome,” embryos. When distinguishing between aneuploidies involving multiple chromosomes—complex aneuploidies—and euploidies, STORK‑A attains an accuracy of 77.6%. The researchers hope to eventually leverage AI and computer‑vision technologies to predict aneuploidy in a fully noninvasive manner.
The new algorithm represents progress in reducing risks associated with IVF embryo selection, minimizing subjectivity, lowering costs, and improving accuracy. Researchers describe it as a compelling example of how AI has the potential to transform medicine.
Next:
Address
Rooms 601 and 602, Building 13, Linhai Intelligent Manufacturing Industrial Park, Jiangnan Subdistrict, Linhai City, Taizhou, Zhejiang Province
Main products
Can’t find the product you’re looking for? Contact us!
If you would like to inquire about our products, please leave your contact information, and we will get in touch with you within 24 hours.
