Background and objective: Artificial intelligence (AI) has emerged as a promising tool for clinical decision support in reproductive medicine, yet the performance of general-purpose large language models (LLMs) in predicting in vitro fertilization (IVF) outcomes remains insufficiently characterized. This exploratory proof-of-concept study aimed to evaluate and compare the out-of-the-box performance of three widely accessible LLM-based systems (ChatGPT, DeepSeek, and Gemini) in forecasting key clinical and laboratory outcomes of IVF treatments. Methods: This retrospective single-center study used data from 1473 autologous IVF/ICSI cycles, each representing a unique patient. For each cycle, relevant clinical and laboratory variables were incorporated into a standardized anonymized patient-level vignette and submitted via the publicly available web interfaces of three LLMs (ChatGPT, DeepSeek, Gemini) without any fine-tuning or internal customization. The models were asked to predict stimulation protocol, ovulation trigger type, total and mature oocyte counts, usable embryo counts, and clinical pregnancy. Predictive performance was evaluated using accuracy and tolerance-based accuracy for categorical and count-based outcomes, mean absolute error for numerical predictions, and the area under the receiver operating characteristic (ROC) curve for clinical pregnancy. Results: Gemini achieved the highest accuracy in predicting stimulation protocols (51.26%) and embryo counts (68.22%), while DeepSeek demonstrated the lowest numerical error for oocyte count predictions. Clinical pregnancy prediction was the most challenging task; all models showed only moderate discrimination, with Gemini achieving the highest AUC (0.711), followed by ChatGPT (0.690) and DeepSeek (0.676). Overall, model performance varied considerably across tasks and remained below thresholds that would be considered sufficient for reliable stand-alone clinical use. Conclusions: In this exploratory proof-of-concept setting, general-purpose AI systems showed variable and overall suboptimal performance in predicting IVF outcomes from standardized clinical vignettes. Although certain models demonstrated relative strengths in specific tasks, none reached the reliability, consistency, or interpretability required for safe clinical implementation. These findings indicate that, in their current form, such models should not be used as clinical decision-support tools for IVF decision-making and that their use should remain restricted to carefully controlled research settings until they have been prospectively validated in multicenter cohorts and systematically compared with rigorously developed, task-specific prediction models. This study provides comparative insight into how these AI systems behave in IVF-related prediction tasks and underscores the need for cautious interpretation of AI-generated outputs.

Background: The aim of this study is to compare the concordance of Preoperative Magnetic Resonance Imaging (MRI) and Intraoperative Frozen Section Analysis, widely used worldwide for Endometrial Cancer (EC), with final pathology results, to calculate their sensitivity and specificity, and to evaluate their diagnostic agreement with final pathology results. Positive predictive values for both MRI and frozen section analysis will also be calculated. Methods: In this retrospective cohort study, patients diagnosed with Endometrioid-Type Endometrial Cancer at the Gynecologic Oncology Surgery Department of Akdeniz University Hospital between January 2020 and December 2023 underwent preoperative MRI to assess the depth of myometrial invasion and intraoperative frozen section examination for surgical staging. The results of both methods were compared with the final pathology reports. Results: A total of 88 patients were included in the study. Patient ages ranged from 34 to 80 years, with a mean age of 57.57 years (SD: 9.65), predominantly in the middle-aged and older population. In the assessment of myometrial invasion depth, MRI demonstrated a sensitivity of 81.6% and a specificity of 88%, while frozen section analysis showed a sensitivity of 73.7% and a specificity of 98.0%. Conclusions: In our study, preoperative MRI demonstrated similar sensitivity and specificity to intraoperative frozen section analysis in determining the depth of myometrial invasion in cases of Endometrioid-Type Endometrial Cancer. Therefore, when intraoperative frozen section analysis is not available, MRI findings may assist surgical planning, particularly in centers where frozen section is limited.

This prospective study evaluated ChatGPT-4 as a decision-support tool by comparing its treatment recommendations with clinical decisions for 87 women (aged 40–65 years) scheduled for hysterectomy. Demographic, clinical, and ultrasonographic data were standardized and submitted to the GPT-4 model, which generated evidence-based treatment suggestions. ChatGPT-4 recommended hysterectomy in 70.1% of cases, aligning with the original clinical decision, and suggested alternatives such as myomectomy (10.3%), hysteroscopy (8.0%), or medical therapy (4.6%) in others. These alternative options were retrospectively judged as appropriate in selected scenarios. Although the model demonstrated guideline-consistent reasoning, it lacked access to imaging, laboratory results, and physical examination findings. The study’s single-center design, absence of sample size calculation, and purely descriptive nature further limit generalizability. Large language models may complement clinical decision-making but should not replace physician expertise. Multicenter studies are needed to validate their reliability and clinical applicability.

Purpose: To compare maternal serum sex hormone-binding globulin (SHBG) and androgen markers in intrauterine growth restriction (IUGR) and gestational-age-matched controls, and to assess their associations with fetal growth. Methods: This prospective case–control study included 45 IUGR pregnancies and 50 controls. Blood was collected in the late third trimester, preferentially antepartum; otherwise within 24 h postpartum (IUGR: 36/45 antepartum, 9/45 postpartum; controls: 9/50 antepartum, 41/50 postpartum). Total testosterone, dehydroepiandrosterone sulfate, estradiol, androstenedione, and SHBG were measured by immunoassay. FAI was calculated as (testosterone [nmol/L]/SHBG [nmol/L]) × 100 after conversion of testosterone from ng/mL to nmol/L. Sensitivity analyses compared hormone levels within antepartum and postpartum subgroups. Exploratory multivariable linear regression in the IUGR group assessed whether SHBG and total testosterone were independently associated with birth weight percentile and abdominal circumference percentile after adjustment for maternal age, systolic blood pressure, and sampling time. Results: IUGR mothers were older, delivered earlier, and had higher systolic blood pressure than controls. Birth weight was lower in IUGR (2190 ± 624 g vs 3306 ± 391 g). SHBG was lower in IUGR (257 ± 117 vs 382 ± 106 nmol/L, p < 0.001), whereas FAI was modestly higher (1.27 ± 1.06 vs 0.85 ± 0.71, p = 0.018). In the antepartum-only subgroup, SHBG remained lower in IUGR, whereas FAI did not differ significantly. Unadjusted correlations of SHBG and total testosterone with fetal growth parameters in IUGR were attenuated after adjustment. Conclusion: IUGR was associated with lower maternal SHBG. The FAI difference was modest and not robust in the antepartum-only analysis. Hormone-growth associations were exploratory and were not independently significant after adjustment.