Annual publication volume fluctuations serve as a significant indicator of the development pace and scholarly interest within a field, making it a key metric for assessing the academic landscape. From 1999 to 2024, the annual publication volume of AI applications in obstetrics and gynecology has shown a consistent upward trend. Over the past five years (2020–2024), the average annual publication volume exceeded 100, reflecting increased scholarly attention and indicating a promising future for AI applications in this field. China, the United States, the United Kingdom, Canada, and South Korea are the leading countries in scientific research output on AI applications in obstetrics and gynecology, with established collaborative networks among them. Universities and hospitals are the primary institutions contributing to this domain. Among the top ten most productive institutions by publication volume, four are based in the United States, while two each are from China and Italy, signaling a strong commitment to advancing research in this area by institutions in these countries. The journal distribution pattern reveals that a significant body of research on AI applications in obstetrics and gynecology has been published across journals in obstetrics and gynecology, oncology, reproductive medicine, multidisciplinary sciences, and medical imaging. The American Journal of Obstetrics and Gynecology stands out as the leading publication in terms of article volume, impact factor, and CiteScore within this research domain. An analysis of publication output and collaboration patterns reveals the absence of dominant core authors in this research field, with inter-team collaboration still underdeveloped. Among the contributors, Laios, Alexandros stands out as the most prolific author. His recent work primarily explores the application of natural language processing (NLP) to predict intraoperative and postoperative outcomes in advanced-stage epithelial ovarian cancer cytoreduction (aEOC) [12], AI-driven predictions of postoperative hospitalization duration and survival rates in patients with high-grade serous ovarian cancer (HGSOC) [13], and AI-based predictions of factors affecting surgical complexity in advanced epithelial ovarian cancer (EOC) [14].

Analysis of high-frequency keywords, co-occurrence patterns, and clustering indicates that the primary areas of focus in current research are AI-assisted diagnosis and treatment, AI-driven health management, and robotic surgery in obstetrics and gynecology. The future trajectory of research is expected to remain centered on AI-driven diagnostic and therapeutic approaches.

In the early stages of medical information system development, AI was primarily applied to electronic health records. However, with the advent of the big data era, the availability of large-scale datasets has enabled the integration of ML and DL techniques in healthcare. AI’s ability to rapidly and accurately recognize and analyze images offers significant advantages in the screening and diagnosis of gynecological tumors and prenatal assessments. Notably, the interpretation of cervical cytology results remains highly subjective, with low inter-pathologist consistency [15]. AI has demonstrated a 90% accuracy rate in cervical cell pathological analysis, performing on par with experienced pathologists [16]. AI algorithms can enhance lesion segmentation accuracy in cervical cancer colposcopy images up to 96% [17]. Furthermore, AI-assisted systems improve pathologists ’ diagnostic accuracy for serous tubal intraepithelial carcinoma (STIC) by 10% and reduce image interpretation time by one-third [18]. The integration of AI technology effectively addresses the challenges posed by the critical shortage and varied expertise levels among pathologists. Tanos et al. [19] developed a Gynecological AL Diagnostics (GAID) model, which achieved an 87% accuracy rate in diagnosing gynecological diseases, covering sub-specialties such as abnormal uterine bleeding, gynecologic endocrinology, gynecologic oncology, pelvic pain disorders, urogynecology, sexually transmitted infections, and vulvoscopic conditions. In prenatal screening, AI systems have significantly improved prediction accuracy for fetal genetic disorders. AI-enhanced ultrasound screening has a 96% detection rate for Down syndrome, though still Lower than the 99% accuracy of non-invasive prenatal DNA testing [20]. AI’s ability to process extensive clinical datasets, integrate information from various sources, and identify hidden patterns associated with disease risks enhances the accuracy of risk prediction for disease onset. As AI technology advances, models can refine their performance through self-optimization, improving their predictive capabilities. A predictive model incorporating age and 51 laboratory biomarkers has shown high accuracy in forecasting ovarian cancer risk, achieving an area under the receiver-operating characteristic curve (AUC) of 0.95 in an internal validation cohort and AUCs of 0.882 and 0.884 in two external validation cohorts [21]. Additionally, AI achieves 85% accuracy in predicting preterm birth and 88% in forecasting gestational hypertension, thus improving the precision of individualized risk assessments [22]. AI also aids in predicting chemotherapy responses, identifying patients most likely to benefit from treatment. Elfiky et al. [23] conducted a retrospective cohort study of 26,946 oncology patients, employing ML to develop a predictive model for estimating short-term mortality risk (30-day and 180-day) at the initiation of chemotherapy. For patients receiving systemic chemotherapy, predicting 30-day mortality risk could minimize treatment-related adverse outcomes and improve patient quality of life [24]. Desbois et al. [25] developed ML methodologies that integrate transcriptomic data and digital pathology for molecular classification and immune phenotype characterization in ovarian cancer. Their findings suggest that transforming growth factor-beta (TGF-β) plays a key role in T-cell exclusion, proposing that targeting TGF-β may enhance the efficacy of immunotherapy in ovarian cancer.

Health management refers to a systematic process that involves comprehensive monitoring, assessment, and intervention aimed at improving health outcomes for individuals. Traditional health management models primarily rely on clinical expertise and human resources, focusing on diagnosing and treating existing medical conditions. However, with the rapid advancement of AI technologies, particularly large language models (LLMs), AI has significantly transformed the field. Its impact now spans various aspects of health management, including disease prediction, personalized health planning, continuous health monitoring, chronic disease management, and optimization of epidemiological research and healthcare policies [26]. The integration of AI addresses key limitations of traditional health management by improving operational efficiency, enabling personalized interventions, optimizing data utilization, and enhancing predictive capabilities [27]. In obstetrics and gynecology, women’s health needs extend across the entire lifespan, including health management from infancy through adolescence, reproductive and fertility care during the childbearing years, preventive strategies during perimenopause, and targeted interventions in older age. AI has the potential to significantly enhance this holistic approach to women’s health. By utilizing AI-powered big data analytics and personalized assessment tools, gynecologists can more accurately monitor patient health, proactively identify emerging risks, and implement effective, life-cycle-oriented health management strategies. AI-driven applications can monitor menstrual cycles and identify potential health issues, such as polycystic ovary syndrome (PCOS), by analyzing symptoms and cycle irregularities. Automated alerts can prompt users to seek medical consultation when clinically necessary. Women of reproductive age can use wearable devices and digital health platforms to track physiological parameters, such as basal body temperature, heart rate, and sleep patterns, facilitating the prediction of menstrual cycles, ovulation windows, and fertility status [28,29,30,31]. These technologies support women aiming to conceive by enabling data-driven family planning and improving preconception care. Additionally, wearable flexible nano-biosensors enable non-invasive, real-time monitoring of the sex hormone estradiol through sweat analysis, offering a promising method for continuous physiological tracking [32]. AI applications not only empower women to take a proactive role in self-management of their health but also enhance clinical decision-making. By leveraging personalized, data-driven methodologies, AI improves treatment efficacy and facilitates comprehensive health management, from preventive care to targeted interventions.

In 1988, Mettler et al. [33] performed the first gynecological procedure using the AESOP robotic system. In 2005, the U.S. Food and Drug Administration (FDA) approved the Da Vinci robotic surgical platform for gynecologic oncology procedures. A year later, Elliott et al. [34] reported the first series of 30 patients who underwent robot-assisted sacrocolpopexy, with follow-up assessments performed 2 Years postoperatively, demonstrating favorable outcomes. In 2007, Bocca et al. [35] documented the world's first robotic-assisted conservative myomectomy, followed by successful conception and full-term delivery. Accumulated evidence suggests that Da Vinci robot-assisted total hysterectomy offers significantly improved perioperative outcomes and enhanced postoperative recovery, particularly for patients with elevated body mass index (BMI) or larger uterine sizes compared to the conventional laparoscopic techniques [36,37,38,39].

Currently, the Da Vinci robotic surgical system is the most widely adopted platform in obstetrics and gynecology, with applications spanning both benign and malignant gynecological conditions. Common benign indications include uterine fibroids, abnormal uterine bleeding, endometriosis, and pelvic organ prolapse, while malignant applications focus on the management of endometrial, cervical, and ovarian cancers. The robotic surgical system provides several technical advantages, such as a high-definition three-dimensional magnified operative view, wristed instrumentation with enhanced dexterity, and tremor filtration. These features allow for precise and stable surgical manipulation within confined anatomical spaces, enabling surgeons to more accurately identify and dissect vascular structures and delicate tissue planes during complex procedures [40]. Robotic surgical systems have established indications for managing benign gynecological conditions, including hysterectomy [41], myomectomy [42], sacrocolpopexy [43], tubal anastomosis [44], endometriosis-related interventions [45], cervical cerclage [46], and robotic-assisted laparo-endoscopic single-site surgery (R-LESS) [47]. However, the clinical value of robotic surgery in managing benign gynecological conditions remains debated. Some researchers argue that robotic procedures are associated with higher costs and prolonged operative times compared to the conventional surgical methods [48]. Currently, robotic surgical systems are most commonly used in the treatment of malignant neoplasms, such as cervical, endometrial, and ovarian cancers. Since the initial report of robot-assisted radical hysterectomy in 2005, the use of robotic platforms for cervical cancer management has expanded globally. Evidence suggests that, compared to open surgery, minimally invasive techniques like laparoscopy or robotic-assisted surgery result in reduced intraoperative blood loss, lower complication rates, shorter hospital stays, and comparable oncologic outcomes, with no significant differences in recurrence rates or mortality between robot-assisted radical hysterectomy and traditional open procedures [49,50,51,52]. Although robotic applications in obstetrics are less prevalent than in gynecology, several significant implementations have been reported. Sayols et al. [53] introduced a robotic system designed to assist surgeons in performing remote operations during anastomosis positioning, coagulation, and placental surface examination. The research team also developed and validated the functional performance of this robot-assisted platform. The system features image stabilization and precise localization of anatomical regions, enhancing intraoperative photocoagulation accuracy, facilitating accurate surgical navigation, and ultimately reducing operative duration. Ahmad et al. [54] developed a compact robotic system aimed at assisting surgeons in performing procedures with enhanced speed and efficiency. Currently, these robotic systems remain in the experimental phase and have not yet been implemented in clinical settings. Additionally, existing physical simulation platforms used in obstetric training lack the ability to replicate dynamic cervical compliance, thus failing to accurately simulate the physiological process of cervical ripening. Luk et al. [55] presented a novel robotic system capable of simulating cervical ripening during the latent phase of labor. This soft robotic device shows promise as a high-fidelity training simulator for obstetric delivery scenarios.

Robot-assisted surgery has evolved into a rapidly advancing field with substantial growth potential. Evidence from current studies indicates that this technology is not only feasible but also capable of delivering effective therapeutic outcomes for carefully selected patient populations [56, 57]. The future progression of robotic systems is likely to be driven by advancements in both hardware and software. Emerging trends suggest a shift toward miniaturized surgical instruments, portable cart platforms, and the integration of real-time tissue feedback with radiographic imaging and AI across a variety of clinical applications.