Couples pursuing fertility treatment deserve realistic expectations. Treatment with the methods of assisted reproductive technology (ART) is time-consuming, costly, and emotionally demanding. That makes medical counselling all the more important, counselling that reflects a couple's individual situation rather than relying on general registry data alone. This is exactly where the Fertility Navigator comes in: an AI software that calculates a personalized pregnancy prediction before treatment begins.
The Fertility Navigator was developed by NEXUS / MEDITEX c/o Johann Reiter in Regensburg together with Prof. Dr. med. habil. Monika Bals-Pratsch and further specialist colleagues. The results have just been published in the Journal für Reproduktionsmedizin und Endokrinologie (2026; 23(2): 64-71). For us at MedITEX, this project is notable for one particular reason: the data foundation comes from the MedITEX database. It demonstrates the scientific potential that lies within consistent, structured treatment documentation.
Why personalized prognoses have been missing
For counselling in daily practice, general registry data and study analyses have so far been the main reference. These are valuable, but they provide success rates for cohorts, not for the individual case. Whether it is the first or a subsequent treatment cycle is not taken into account. This can distort the individual prognosis, resulting in a probability of pregnancy that is either too low or too high.
Classical statistical methods reach their limits here. The biological and individual preconditions of a couple are highly complex, and the choice of treatment plan has so far depended heavily on the clinical experience of the treating physicians. Modern methods of artificial intelligence can capture this complexity far better, provided the right data foundation is available.
A broad data foundation
Machine learning requires large, high-quality datasets. These are precisely what the legally mandated documentation of ART cycles provides, documentation that has been carried out digitally and prospectively in Germany for almost 30 years.
For the development of the Fertility Navigator, a subset of the MedITEX database was fully anonymized. The data foundation comprises several tens of thousands of cycles from several IVF centers over a period of roughly a decade. The large majority of cycles was used to train the models, with a smaller portion reserved for independent testing and validation. One independent center served as the validation center to reduce bias between centers.
Only pretreatment diagnostic data were analyzedi, in two levels of granularity. The first group comprises fertility-relevant medical history data that couples can provide themselves, such as age, weight, medical history, or smoking status. The second group adds laboratory data, namely the woman's hormone parameters (AMH, FSH, LH, estradiol, progesterone) and the man's spermiogram parameters. Predictive accuracy increases with the availability of additional relevant data.
An ensemble of five AI models
The Fertility Navigator is not a single model but an ensemble of five selected AI models. It draws on established machine learning algorithms such as Random Forest, Gradient Boosting, and XGBoost, as well as deep learning based on neural networks. The advantage of this combination lies in its stability: if one algorithm responds incorrectly to a particular input, the others can compensate. The ensemble-based aggregation of the individual predictions therefore increases robustness and reliability.
The result is expressed as a rating from 0 to 5that can be converted directly into percentage success rates, ranging from verylow to very high probability. This is clinically relevant, because successrates are usually discussed as percentages in the medical counselling session.The chances are calculated for the first three treatment cycles, includingcumulatively.
What the results show
Performance was assessed using the established metrics AUROC and F1 score. The values reveal a clear trend: predictive accuracy rises both with the number of available features and with the number of cycles considered. Including hormone and spermiogram values improved the AUROC to up to 0.65 and the F1 score to up to 0.60. The Fertility Navigator achieves its highest accuracy when a cumulative prognosis over three cycles is produced on the basis of complete medical history and laboratory data.
The analysis also made clear how decisive data quality is. When checking the BMI and obesity fields, input errors appeared in 8.3 to 26.8 percent of cases. For the reliability of any AI-based prognosis, the principle holds: it is only ever as good as the quality of the data entered. Responsibility for this rests with the centers as the point of data entry, supported by documentation software that makes structured and complete recording easier.
Particularly valuable for couples with an unfavorable prognosis
The real clinical benefit becomes apparent precisely where reliable guidance has been lacking. For couples with a good prognosis, orientation values based on an ideal patient are available as an aid. For couples with a poor prognosis there is no comparable calculation, because this group is very heterogeneous and case numbers are small.
This is exactly where the Fertility Navigator can add objectivity. It identifies couples with a good prognosis, but inparticular also those with an unfavorable one, such as women aged 40 and over or younger patients with premature ovarian insufficiency. A realistic individual assessment helps to avoid cycles with little prospect of success and, where the prognosis is favorable, to recommend treatment in a targeted way. Given costs of around 4,000 euros per cycle, which fall entirely on the couple when reimbursement is not available, this is significant not only medically but also economically.
The framing matters here: the Fertility Navigator is not intended to replace the medical counselling session before ART, but to complement it with an objectifying individual prognosis tool.
Outlook: from prognosis to personalized therapy
According to the authors, the Fertility Navigator is one of the first approaches in fertility treatment to calculate a data-based pregnancy prognosis from an individual's pretreatment diagnostics. Its current potential is relevant, but there is room to grow. The next logical step is a broader data foundation.
This is where a major opportunity for reproductive medicine lies. Using the data of all MedITEX centers would allow predictive quality to be refined further on a considerably larger basis. Looking ahead, AI will personalize not only the prognosis but also the therapy, for example in selecting the stimulation protocol and suitable medications.
For us at MedITEX, this project confirms a fundamental conviction: carefully documented treatment data are more than a legal obligation. They are the foundation on which the next generation of personalized fertility medicine is built.
Source and full publication: Bals-Pratsch, M; Reiter, J; Schindler, M; Murr, A; Tinneberg, HR (2026). Der Fertilitätsnavigator als personalisiertes Schwangerschaftsvorhersagemodell bei der Assistierten Reproduktion (ART). J Reproduktionsmed Endokrinol, 23(2): 64-71.
