Introduction:
Taking precautionary measures is of particular importance in midwifery allowing midwives as well as pregnant women to prevent unwanted and difficult situations. The AI-based digital twins technology can be used to represent personalised knowledge pertaining to a woman’s medical, phycological, social environment status, thus facilitating preventive problem solving through a pregnancy period. This research aims at developing digital twins for pregnant women in order to simulate women profiles, investigate scenarios regarding the development of a pregnancy and assist in suggesting taking the necessary measures. Additionally, the digital twins models can be used for educational purposes allowing midwifery students to evaluate their knowledge and calibrate their skills.

Material and Methods:
This study collects data from a group of 600 pregnant women in Greece. The data was collected in two phases. The group of women was randomly split into two groups. Then, phase 1 was carried out through semi-structured interviews that allowed group 1 women to express their expectation, concerns, problems, etc., regarding their pregnancy. During phase 1, a group of 35 were interviewed. Phase 2 involved the administration of a questionnaire which was developed considering the results of the interviews and it was delivered to the group 2 women. The study satisfies the requirements of GDPR, thus the women who agreed to participate in the study had earlier provided their consensus for collecting and analyzing their data. Fuzzy logic and clustering methods were used to analyse the data and develop the digital twins. Scenarios investigated and analyzed using the digital twins were evaluated for their credibility, by calculating accuracy measures as well as by consulting field experts, such as gynecologists.

Results:
The results identify several features, such as macrosomia, depression, Body Mass Index that contribute to developing a highly personalized digital twin model. Aspects of the medical, social and phycological dimensions of the pregnant women profile are extracted and used for developing a digital twin to predict risks e.g., related to pre-eclampsia, anxiety, etc. Results indicate the potential of using digital twins for preventive problem solving as well as for educational reasons.

Conclusions:
The proliferation of AI applications in many other domains spawn new opportunities for applying AI innovative methods in midwifery. This study suggests that digital twins represent a promising technology for facilitating several aspects pertaining to pregnancy such as supporting risk assessment, women profile assessment, students training, etc.