Automated follicle measurement and microvascular imaging accuracy in assessing diminished ovarian reserve via 3D ultrasound

Purpose This study aimed to investigate the diagnostic values of automated follicle measurement and superb microvascular imaging using transvaginal three-dimensional (3D) ultrasound for diminished ovarian reserve (DOR) in women of reproductive age. Methods This retrospective study included women who underwent ovarian reserve function assessment at the Affiliated Obstetrics and Gynecology Hospital of Tongji University between November 2022 and June 2023. The participants were categorized into the DOR and normal groups based on the anti-Müllerian hormone (AMH) and follicle-stimulating hormone (FSH) levels. Results This study included 109 participants: 36 with diminished ovarian reserve (DOR) and 73 with normal ovarian function. Significant differences emerged between the groups in age, body mass index, AMH, FSH, the FSH/LH ratio, and estradiol (all p < 0.05). Transvaginal 3D ultrasound parameters also differed significantly, including antral follicle count (AFC), total ovarian volume (OV), the stroma-to-antrum ratio (S/A), vascularization index (VI), flow index (FI), vascularization-flow index (VFI), and ovarian interstitial artery peak and end-diastolic velocities (PSV, EDV) (all p < 0.05). The efficacy of the ultrasound–serum combined index in diagnosing DOR (AUC = 0.995, 95% CI: 0.949–1.000) is superior to that of the multi-parameter ultrasound index (AUC = 0.973, 95% CI: 0.914–0.996), AFC (AUC = 0.965, 95% CI: 0.902–0.993), and AMH (AUC = 0.974, 95% CI: 0.916–0.996). According to the ROC curve analysis, when selecting the cutoff value based on the highest sensitivity, the optimal cutoff for AFC was 6.5. After including the age variable, least absolute shrinkage and selection operator (LASSO) regression identified AFC, stromal OV, total VI, total FI, and EDV as optimal variables. Among them, AFC had the highest weight coefficient. The OR values of AFC (OR = 0.596, 95% CI: 0.442–0.804, p < 0.001) and total FI (OR = 0.853, 95% CI: 0.737–0.988, p = 0.033) in the multivariate analysis are statistically significant. Decision curve analysis (DCA) demonstrated that the PRE_LASSO model consistently outperformed the PRE_AFC model in net benefit across most threshold ranges. The scatter plot of net reclassification improvement (NRI) showed that most case patients had increased risk, while control patients had reduced risk, indicating that the PRE_LASSO model exhibits superior clinical application value and risk stratification capability. Conclusion This study identified AFC as the key predictor factor for DOR, with an AUC of 0.965 (95% CI: 0.902–0.993), demonstrating high diagnostic value. These results underscore the promising potential of transvaginal 3D ultrasound techniques to enhance the precision of ovarian reserve assessments.