Senior Embryologist, ESHRE Member: Iulia Raica
In vitro fertilization (IVF) remains one of the most effective treatments for infertility. However, success rates are still limited by biological variability and multiple clinical factors.
Embryo selection represents the critical stage of IVF, as only a small proportion of embryos have real implantation potential and the ability to develop into a live birth. The goal of embryo selection is to identify the most viable embryos from a larger cohort of fertilized oocytes, many of which are non-viable due to abnormal development or chromosomal imbalances.
Embryo implantation depends on several factors:
- Embryo quality
- Endometrial receptivity
- Hormonal balance
- Lifestyle factors
Among these, embryo quality remains the primary determinant of IVF success.
The Biological Limits of IVF Success
One of the most significant limiting factors in IVF is embryonic aneuploidy, strongly associated with maternal age.
In addition to chromosomal abnormalities, cytoplasmic defects in oocytes — including organelle dysfunction, altered gene expression, and lack of essential macromolecules — can significantly compromise embryo development, even when chromosomal status appears normal.
Structural chromosomal abnormalities such as Robertsonian translocations or inversions are another cause of infertility. These often present with a normal phenotype and are detected only through karyotyping. Y chromosome microdeletions in sperm can also reduce euploidy rates.
Understanding chromosomal defects is therefore essential in comprehensive infertility assessment. It helps patients understand their diagnosis and enables physicians to design targeted treatment strategies.
Artificial Intelligence in Reproductive Medicine
In this context, the integration of artificial intelligence (AI) into reproductive medicine offers powerful analytical and decision-support tools capable of optimizing clinical outcomes and standardizing laboratory processes.
Deep learning algorithms can automatically analyze embryo images and identify patterns associated with embryo viability — patterns that are often inaccessible to human visual assessment.
AI applications in IVF now include:
- Oocyte evaluation
- Sperm assessment
- ICSI optimization
- Time-lapse embryo monitoring
- Detection of abnormal fertilization
- Blastocyst ranking
- Non-invasive ploidy prediction
Some AI platforms demonstrate performance comparable to non-invasive genetic testing, with the future potential of estimating embryo ploidy without biopsy.
Why Embryo Selection Is Critical in IVF
Only a small percentage of embryos generated in IVF cycles have real implantation potential. Most embryos are excluded due to abnormal development or chromosomal aneuploidy.
Traditional embryo evaluation relies heavily on visual observation and embryologist experience. While valuable, this method has limitations:
- Subjectivity
- Inter-observer variability
- Difficulty detecting subtle abnormalities
- Limited standardization
Technological advancements now allow more objective and reproducible assessment methods.
AI in Oocyte Evaluation
Oocyte quality is fundamental for IVF success.
Using convolutional neural networks (CNNs), AI can assess:
- Oocyte texture and structure
- Zona pellucida quality
- Organelle organization
These systems provide more consistent and precise evaluations compared to manual assessment.
AI in ICSI (Intracytoplasmic Sperm Injection)
ICSI is a delicate procedure requiring high precision.
AI systems can:
- Identify optimal oocyte orientation
- Determine ideal injection points
- Achieve precision rates up to 99%
This reduces procedural errors and increases efficiency.
Detection of Abnormal Fertilization
After fertilization, normal embryos display two pronuclei. Abnormal embryos (0, 1, or 3 pronuclei) can be identified by AI systems with over 93% accuracy, allowing early exclusion of non-viable embryos.
Non-Invasive Ploidy Prediction
Preimplantation genetic testing (PGT-A) is effective but invasive and costly.
AI systems analyze time-lapse embryo images and detect patterns associated with aneuploidy, potentially reducing the need for embryo biopsy.
Platforms such as STORK and BELA demonstrate promising results.
Blastocyst Evaluation and Live Birth Prediction
At day 5–6, embryos reach the blastocyst stage. AI platforms like STORK (≈96% accuracy) and AIRE/ERICA (70–77% accuracy) analyze embryo images and clinical data to predict implantation success and live birth probability.
Broader AI Applications in IVF
AI is also used for:
- Sperm selection and motility analysis
- Ovarian stimulation personalization
- Laboratory workflow optimization
- Culture environment monitoring
These applications reduce time, cost, and variability.
Adoption Trends
Global surveys show a clear increase in AI use in IVF:
- 24.8% adoption in 2022
- Over 53% regular or occasional use in 2025
Over 80% of specialists plan to invest in AI within the next five years.
Ethical and Practical Challenges
Despite its benefits, AI implementation raises concerns:
- High costs
- Limited training
- Regulatory gaps
- Ethical considerations in embryo selection
AI must remain a decision-support tool — not a replacement for medical expertise.
Conclusion
Artificial intelligence has transformative potential in IVF by increasing precision, objectivity, and efficiency.While not a miracle solution, AI represents a major step toward personalized, accessible, and optimized reproductive medicine. Its success depends on balancing technological innovation with ethical responsibility and human expertise.
Embr. Iulia Raica
ESHRE Member, OBBCBCSSR Member, Specialization: Fertility Center Kinderwunschzentrum GoldenessKreutz Private Clinic, Vienna, Austria; ESHRE Course Choosing and using embryos: novel approaches to embryo selection, Krakow, ESHRE PGT Consortium SeminarIn vitro fertilisation (IVF)
A fertility treatment in which eggs are fertilised with sperm in a laboratory before an embryo is transferred into the uterus.
Artificial intelligence (AI)
Computer-based technology that analyses large amounts of data to identify patterns, generate estimates and support clinical or laboratory decisions.
Deep learning
A form of artificial intelligence that uses multilayered algorithms to identify complex patterns in images and other data.
Convolutional neural network (CNN)
A deep-learning model designed primarily to analyse visual information. In reproductive medicine, it may be used to assess images of oocytes and embryos.
Oocyte
An immature or mature female reproductive cell, commonly referred to as an egg.
Embryo selection
The process of evaluating embryos to determine which may have the greatest potential for implantation and continued development.
Embryo viability
The potential of an embryo to continue developing and possibly implant in the uterus. Viability assessments provide estimates and cannot guarantee pregnancy or live birth.
Aneuploidy
The presence of an abnormal number of chromosomes in a cell or embryo. Embryonic aneuploidy may reduce implantation potential and increase the risk of miscarriage.
Euploid embryo
An embryo with the expected number of chromosomes. A euploid result does not guarantee implantation, pregnancy or live birth.
Karyotyping
A laboratory examination used to assess the number and structure of chromosomes and identify certain chromosomal abnormalities.
ICSI (Intracytoplasmic Sperm Injection)
An assisted reproduction technique in which a single sperm cell is injected directly into an oocyte to achieve fertilisation.
Pronuclei
The nuclei originating from the egg and sperm that are visible shortly after fertilisation. Normally fertilised oocytes typically contain two pronuclei.
Time-lapse embryo monitoring
A laboratory technique that captures images of developing embryos at regular intervals without removing them from the controlled incubator environment.
PGT-A (Preimplantation Genetic Testing for Aneuploidy)
A genetic test performed on cells collected from an embryo to assess whether it has the expected number of chromosomes before embryo transfer.
Embryo ploidy
The number of complete sets of chromosomes present in an embryo. Ploidy assessment helps identify whether an embryo is euploid or has a chromosomal imbalance.
Non-invasive ploidy prediction
The use of embryo images, developmental data or artificial intelligence to estimate chromosomal status without performing an embryo biopsy. It provides a probability rather than a definitive genetic diagnosis.
Blastocyst
An embryo that has developed for approximately five to six days after fertilisation and may be considered for transfer, genetic testing or freezing.
Endometrial receptivity
The ability of the uterine lining to support embryo attachment and implantation during the appropriate stage of the menstrual cycle.
AI-assisted embryo assessment
The use of artificial intelligence to analyse embryo images and developmental data. It supports embryologists by providing structured information but does not replace their expertise or guarantee treatment success.
