| Bals-Pratsch M et al. | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
Der Fertilitätsnavigator als personalisiertes Schwangerschaftsvorhersagemodell bei der Assistierten Reproduktion (ART) // The Fertility Navigator as a personalized pregnancy prediction model in assisted reproductive technology (ART) Journal für Reproduktionsmedizin und Endokrinologie - Journal of Reproductive Medicine and Endocrinology 2026; 23 (2): 64-71 Volltext (HTML) Summary Praxisrelevanz Abbildungen Keywords: big data, Fertilitätsnavigator, KI, künstliche Intelligenz, maschinelles Lernen, ML, personalisierte assistierte Reproduktion, personalisierte Behandlungsstrategien, Vorhersage des Schwangerschaftserfolgs, Vorhersagemodelle, AI, Artificial Intelligence, fertility navigator, Machine Learning, personalized assisted reproduction, personalized treatment strategies, prediction of pregnancy outcome, predictive models Artificial intelligence (AI) enables the development of software tools that can provide personalized predictions and individualized treatment protocols for an ART cycle. Machine learning (ML) models require very large datasets (“Big Data”) for training and testing. Owing to the professional obligation to document treatment data in assisted reproduction (ART) and the prospective digital recording of ART cycles in Germany for almost 30 years (D·I·R database), such datasets are, in principle, available. As a first step, AI models based on comprehensive pretreatment diagnostic data, including medical history and laboratory parameters, were developed (Fertility Navigator, CRITEX GmbH, Regensburg, Germany) to calculate an individualized prediction of pregnancy success before ART treatment. Training and testing (validation) were performed on a dataset of 89,452 cycles. For this purpose, a subset of the MedITEX database (CRITEX GmbH, Regensburg) was fully anonymized. The performance of the Fertility Navigator for personalized pregnancy prediction was assessed using established metrics such as AUROC and F1-score. These findings suggest that AI-based personalized pregnancy prediction may support medical counselling and treatment decisions prior to ART, particularly in couples with a poor prognosis. To improve predictive accuracy further, the Fertility Navigator should be developed and validated on even larger datasets. Suitable Big-Data sources, such as the complete MedITEX dataset or the D·I·R database, are in principle available. Further development of the Fertility Navigator also appears promising with regard to future personalized treatment strategies in ART.
Kurzfassung: Mit den Möglichkeiten künstlicher Intelligenz (KI) können Softwareprogramme entwickelt werden, die personalisierte Prognosen und individualisierte Behandlungsprotokolle für einen ART-Zyklus ermöglichen. Voraussetzung für das maschinelle Lernen (Machine Learning, ML) von KI-Modellen sind vor allem sehr große Datenmengen („Big Data“), um diese zu trainieren und zu testen. |
