AI Image Analysis Shown to Predict Embryo Health

In-vitro fertilization (Elena Kontogianni, Pixabay. https://pixabay.com/photos/ivf-fertility-infertility-1514174/)

28 June 2022. An algorithm designed to analyze microscopic images is shown to largely predict the genetic integrity of embryos from in-vitro fertilization. Findings from a study evaluating the technology developed by Presagen, an artificial intelligence health analytics company in San Francisco, appear in the 8 June issue of the journal Human Reproduction.

Presagen creates analytics based on A.I. to address women’s health. The company says it collaborates with clinics worldwide to design its algorithms, for analytics that represent a wide range of races and ethnic groups, as well as identify potential errors early in their development. For embryo health algorithms, Presagen established Life Whisperer, a subsidiary that works with in-vitro fertilization or IVF clinics to analyze images of early embryos for viability to find the healthiest candidates for implantation. Life Whisperer says it offers a cloud-based service that captures data and returns results quickly to IVF clinics worldwide.

One of the Life Whisperer algorithms analyzes a blastocyst formed a few days after fertilization for chromosome abnormalities. A healthy blastocyst has 46 chromosomes indicating normal genetic development, but those with different numbers, called an aneuploid, indicate a greater chance for developing into a genetic disorder. The current method for discovering genetic abnormality in a blastocyst is the pre-implantation genetic testing for aneuploidy or PGT-A test. That test uses a biopsy to take a tiny sample of cells from a blastocyst for genetic analysis. Presagen says PGT-A tests are invasive and risky to the developing embryo, with additional high costs for genetic testing.

Correlated algorithm to PGT-A results

A team from Presagen developed the Life Whisperer algorithm trained with blastocyst images collected from 10 IVF clinics in the U.S., Spain, India, and Malaysia. Some 5,050 microscopic blastocyst-stage embryo images at day five of development were selected from a pool of more than 15,000 images to train the algorithm. The microscopic images were also linked to PGT-A genetic test metadata. The researchers correlated analytics from the algorithm with PGT-A test results, looking for accuracy in predicting a genetically healthy blastocyst called a euploid, or an aneuploid, a genetically abnormal blastocyst.

The findings show the Life Whisperer algorithm accurately predicted euploid status 77 percent of the time in a blind-test data set, following removal of poor-quality or mislabeled images. In addition, the researchers found a high probability, from 82 to 97 percent, of an embryo with a high score from the A.I. algorithm also being a euploid, with the rank-order of algorithms scores about 26 percent higher than from random ranking. Further analysis shows results from the algorithm generalized well to patient demographics, and could also detect mosaic embryos, where mistakes in cell division occur, leading to miscarriages or birth defects.

Sonya Diakiw, Presagen’s chief medical scientist and lead author of the paper, notes in a company statement that the Life Whisperer algorithm will not likely replace the PGT-A test, but can still be a valuable tool for IVF decisions. “Because this assessment is based on images alone,” says Diakiw, “it is not as accurate as PGT-A itself, which involves actual DNA sequencing.” Diakiw adds however, “PGT-A only tests five cells from a total of around 200, so it is not always representative of the entire embryo. Life Whisperer genetics is a whole-embryo assessment of genetic integrity that does not require any invasive procedures, which can be used to prioritize embryos for use in IVF procedures.”

More from Science & Enterprise:

• Digital Biomarkers Reveal Alzheimer’s Gender Differences
• Challenge Seeks Tech, Models to Reduce Stillbirths
• Trial Set for Ingested Colon Cancer Detection Capsule
• Digital Neuroscience Companies Partner on AR/AI Diagnostics
• NIH Small Biz Grant Supports Uterine Health Diagnostic

We designed Science & Enterprise for busy readers including investors, researchers, entrepreneurs, and students. Except for a narrow cookies and privacy strip for first-time visitors, we have no pop-ups blocking the entire page, nor distracting animated GIF graphics. If you want to subscribe for daily email alerts, you can do that here, or find the link in the upper left-hand corner of the desktop page. The site is free, with no paywall. But, of course, donations are gratefully accepted.

*     *     *