AI Analysis of Endometrial Receptivity in Georgia: Technical Principles and Clinical Applications

When a patient asks: “Is the AI analysis of endometrial receptivity in Georgia really reliable?”

In the assisted reproduction clinic, a 38-year-old patient who had experienced three failed transfers of high-quality embryos pulled up an institutional introduction on her phone, pointed to the phrase “AI analysis of endometrial receptivity” and asked me: “This technology is not even very common in our country. Is it really trustworthy in Georgia? Could it just be a gimmick?”

This question is not an isolated case. As Georgia’s assisted reproductive policies remain relatively stable and its success rate reputation grows, more and more domestic patients are turning their attention to fertility centers in Tbilisi and Batumi. Endometrial receptivity analysis (especially the new AI-based methods) has become an option recommended by many doctors for patients with “recurrent implantation failure.” Below, we break down this issue from the perspective of the technology itself and the actual situation in Georgia.

What is AI analysis of endometrial receptivity?

Traditionally, assessing whether the endometrium is capable of accepting an embryo has relied mainly on two methods:

• Morphological assessment: Observing endometrial thickness, morphology, and blood flow signals via transvaginal ultrasound.
• Gene expression profiling (e.g., ERA): Taking a small piece of endometrial tissue to analyze the expression levels of specific genes and determine the precise timing of the “implantation window.”

AI analysis builds on this by using machine learning models to integrate the patient’s ultrasound images, hormone levels, previous cycle data, and even gene methylation information to directly predict the optimal transfer time or endometrial status score. Currently, the AI tools used by some institutions in Georgia mainly fall into three categories:

In Georgia, the most common application is a combination of “Transcriptomic AI” and “Integrated Model AI.” This is because fertility centers in Georgia generally accept frozen embryo cycles, allowing time for endometrial preparation and analysis.

Why is there a need for “AI analysis of endometrial receptivity”?

Recurrent implantation failure (RIF) has always been a challenge in the field of assisted reproduction. Approximately 15% to 25% of high-quality embryos still fail to implant after transfer, and about two-thirds of these cases are related to abnormal endometrial receptivity. Although traditional ERA has a relatively high accuracy (about 80% to 85%), it has significant limitations:

1. Invasive procedure: Requires hysteroscopic biopsy to remove a small piece of endometrium, which may cause bleeding or infection.
2. Window limitation: ERA can only detect the P+5 to P+7 window corresponding to a single endometrial tissue sample and cannot dynamically assess changes throughout the cycle.
3. High cost: Overseas (including Georgia), the cost of ERA testing is approximately $1,500 to $2,500, and samples need to be sent to laboratories in the US or Europe.

AI analysis attempts to address these issues: by obtaining data through non-invasive methods (such as ultrasound and blood tests) and using algorithms to infer endometrial status, it reduces trauma and may allow for multiple assessments. This is particularly attractive for patients with thin endometrium, a history of intrauterine adhesions, or those who cannot tolerate biopsy.

Practitioner’s observation: The real status of AI analysis in Georgia

Having practiced in Georgia for ten years and directly interacted with the main fertility centers in Tbilisi, Batumi, and Kutaisi, I can objectively describe the following points:

• Not all centers have this technology. Currently, only 2 to 3 institutions that see a high volume of international patients (such as Beta Plus and some branches of Invitro Georgia) have truly integrated AI analysis into their routine workflow. Most small to medium-sized centers still rely on traditional ERA or ultrasound assessment.
• AI models mostly come from European or Israeli partners. Georgia’s local reproductive database is limited, and the model training data primarily comes from European populations. Caution is needed regarding its applicability to Asian populations. Some centers note “model extrapolation degree” in their analysis reports.
• Price is 30% to 40% cheaper than traditional ERA. The cost of AI analysis is approximately $800 to $1,200 (including ultrasound, hormone testing, and algorithm report), but an additional endometrial sampling for gene verification may still require out-of-pocket payment.
• Result interpretation depends on doctor’s experience. AI only provides a “recommended transfer time” or “endometrial receptivity score (0-10).” Whether to adopt it requires the doctor to integrate the patient’s hormone curve, endometrial morphology, and previous embryo grade for a comprehensive judgment.

Practical procedure: How many steps are needed for an AI analysis in Georgia?

Step 1: Screening suitable candidates

Doctors typically recommend it only for the following groups:

• At least 2 failed transfers of high-quality embryos (embryo factors excluded);
• Endometrial thickness and morphology are basically normal but implantation has failed;
• Previous ERA results were normal but implantation still failed, suspected dynamic window shift;
• Unable to tolerate hysteroscopy or endometrial biopsy (e.g., severe intrauterine adhesions, post-operative recovery).

Unsuitable candidates: Patients with very thin endometrium (<5mm), acute endometritis, untreated hydrosalpinx, severe adenomyosis lesions, etc. AI analysis cannot replace the treatment of these underlying conditions.

Step 2: Cycle preparation and data collection

Usually a natural cycle or hormone replacement therapy (HRT) cycle. Starting around day 10 of the menstrual cycle:

• Transvaginal ultrasound: Continuous monitoring of endometrial thickness, morphology, peristaltic waves, and resistance index (RI) of blood flow.
• Hormone testing: E2, P4, LH levels (usually every 2 days).
• Saliva/fingerstick blood sample (optional): Some AI models require DNA methylation or RNA expression, but centers in Georgia mostly use non-invasive methods, only taking fingerstick blood.

Step 3: AI model analysis

All data is uploaded to a secure analysis platform (report usually returned within 24 to 48 hours). The report includes:

• Current cycle endometrial receptivity score (dynamic curve);
• Suggested optimal transfer time point (e.g., P+5.5 days, P+6 days, etc.);
• If the deviation is large, it may also indicate “whether an additional endometrial biopsy for verification is recommended.”

Step 4: Transfer decision

The doctor adjusts the transfer window based on the AI report, combined with the patient’s own LH surge or progesterone administration timing. For example, if the original plan was transfer on P+5 days, but AI suggests P+6.5 days, the doctor may postpone by one day and arrange another ultrasound to confirm endometrial status.

Easily overlooked details and pitfalls

1. AI analysis cannot 100% replace pathological diagnosis. If there are organic lesions such as chronic endometritis (CD138 positive), endometrial tuberculosis, or isthmocele fluid, the AI model may be misled by “pseudo-normal” data. Hysteroscopy or endometrial biopsy must be done first to rule out these underlying issues.

2. Hormonal fluctuations within the cycle can affect model accuracy. Some AI models assume a stable hormone curve, but in some patients (especially those with diminished ovarian reserve or using ovulation induction drugs), hormone levels fluctuate significantly, and the model’s predicted window shift may be inaccurate.

3. Limitations of a “one-time AI analysis.” Endometrial receptivity changes dynamically within a menstrual cycle, but AI analysis usually involves only one data collection (e.g., on day 5 after ovulation). If the window shift is related to hormonal pulses, a single assessment may miss it.

4. Language and communication costs. AI reports in Georgia are mostly in Russian or English. Some Chinese patients need to wait for translation, and doctors often do not explain the algorithm logic in sufficient detail, which can easily lead to misunderstandings. It is recommended that patients complete consultations with the help of a translator or through telemedicine support.

Interpretation of test indicators: How to view AI analysis results

Taking the “EndoPredict” integrated model used by a center in Georgia as an example, the output indicators include:

Note: These cut-off values mostly come from European Caucasian databases. Centers in Georgia are gradually accumulating local population data. After receiving the report, patients should ideally have the doctor recalibrate it based on their personal menstrual cycle length and hormone peak.

Risk reminders

AI analysis of endometrial receptivity in Georgia is still an emerging technology, lacking large-scale prospective controlled studies to confirm its superiority over traditional ERA. Patients need to be aware of the following risks:

• Misjudgment risk: The model may give incorrect window predictions in cases of specific endometrial pathologies (e.g., focal hyperplasia, polyps, adhesions), leading to failed implantation or early miscarriage after transfer.
• Additional costs: If the AI analysis result is not ideal, the doctor may recommend a supplementary ERA, potentially making the total cost higher than ERA alone.
• Time delay: An AI analysis usually requires waiting for 1 to 2 cycles (including data collection and model feedback). If the result needs verification, it may waste another cycle.
• Legal and certification issues: Currently, the Georgian Ministry of Health has not issued specific guidelines for AI-assisted reproductive analysis. Each center chooses its own technology platform, and data privacy protection standards vary. It is recommended to confirm whether the center has ISO 27001 or GDPR-related certifications before consultation.

Suggestions for next steps

If you are considering undergoing AI analysis of endometrial receptivity in Georgia, you can first complete the following checks as planned:

• Basic fertility assessment (AMH, FSH, antral follicle count);
• Hysteroscopy (to rule out polyps, adhesions, chronic endometritis);
• Couple’s chromosome karyotyping and genetic counseling;
• Infectious disease screening and passport/visa preparation.

After confirming that there are no organic endometrial lesions, discuss with your fertility doctor whether to introduce AI analysis. If the number of embryos is limited (e.g., only 1-2 transferable blastocysts), AI analysis may help improve the match of the transfer window; if the number of embryos is larger (≥5), you may prioritize a direct transfer first and only do window testing if it fails.

Final reminder: Currently, institutions in Georgia offering AI endometrial receptivity analysis have not published local population validation data. Before making a decision, it is advisable to ask the center for success rate statistics from previous Asian patients (at least 50 cases) and sign an informed consent form clearly stating the technical limitations.