Follicular sphericity based on three-dimensional transvaginal ultrasound algorithms predicts ovarian responsiveness to in vitro fertilization.

This was a prospective observational study that linked the information of patients from the Electronic Medical Record System (EMRS) in order to evaluate women who commenced their first COS cycle with a gonadotropin-releasing hormone antagonist (GnRH-ant) protocol (using either in vitro fertilization [IVF] or intracytoplasmic sperm injection [ICSI]) between March 2023 and March 2024. Institutional Review Board approval (No.2023-049-01) was obtained and written informed consent was received from all the participants prior to the study. All patients diagnosed with male factor, ovulatory dysfunction, tubal factor, diminished ovarian reserve and unexplained factor, with both ovaries present were included in this study. Patients who had undergone preimplantation genetic testing (PGT) cycles, or possessed male/female chromosomal abnormalities, endometriosis, uterine factor or serious systemic illness, or an abnormal ovarian mass of 3 cm in diameter were excluded from this study. Patients satisfying the eligibility criteria underwent a 3D ultrasound scan while those with incomplete information, without oocyte retrieval, abnormal ovarian mass, early termination of IVF-ET, or low 3D-US image quality dropped out of the study. Our study thus ultimately comprised a total of 248 patients who underwent their first COS cycle with fresh ET (Fig.  1 ). Fig. 1 Patient selection and study protocol Patient selection and study protocol COS was performed using a GnRH-ant protocol as previously described [ 19 ]. The starting dose was determined by antral follicle count (AFC), the woman’s age, BMI, and other factors. A starting dose of 100–149 IU is recommended for patients with a very high ovarian response, 150–225 IU for those with normal ovarian response, and 225–300 IU for those with a poor ovarian response [ 20 ]. This protocol was initiated on day 2 or 3 of the cycle with a dose of recombinant follicle-stimulating hormone (rFSH) ranging between 75 and 300 IU. The patients were administered 0.25 mg of GnRH-ant (Orgalutran, Organon, the Netherlands or Cetrorelix, Merck Serono, Switzerland) daily when at least one of the following criteria was fulfilled: (i) presence of at least one follicle > 14 mm in diameter; (ii) a serum estrogen level > 600 pg/mL; and (iii) a serum luteinizing hormone level > 10 IU/L [ 21 ]. During the follicular monitoring period, the physicians adjusted the Gn dose (≤ 150 IU) fewer than twice according to follicular growth (and commonly given on days 4–7). The adjustment was performed based on the experience of the physician. A Voluson E8 (GE Healthcare Technology, Milwaukee, WI) instrument equipped with a 5- to 9-MHz transvaginal volume probe was used for all ultrasonographic follicular monitoring. Two senior physicians monitored follicle development during COS every 1–2 days, and follicular diameter was recorded after confirming that the measurements were consistent between the two physicians until the day of HCG administration. The mean of the maximal 2D-diameter X and the 2D-diameter Y perpendicular to the X diameter of the follicle (i.e., orthogonal diameters) were considered the 2D diameters of a follicle, and the numbers of follicles in both ovaries were recorded; the calculations were then based on the sum of follicles in the right and left ovaries. For the first four days after Gn administration, follicles greater than 5 mm in diameter were recorded; for days 5–7 after Gn administration, follicles over 8 mm in diameter were recorded; and for the day of HCG administration, follicles greater than 10 mm in diameter were recorded. Ovarian volumes were obtained by 3D-ultrasound on the Gn start day and the HCG day, and the three maximal perpendicular diameters were measured (calipers)—with the calculation of ovarian volume conducted using the formula volume of ovary = D1 × D2 × D3 × Π/6 (cm 3 ) [ 22 ] (Supplement Fig.  1 ). For 3D-US follicle monitoring, we applied the sonography-based Automated Volume Count system (SonoAVC) (GE Medical Systems), a software program that automatically identifies and measures follicles within a 3D volume and increases follicular assessment efficiency. On the HCG trigger day, the plane of the maximal diameter of the ovary was then revealed, and the image was stabilized for 3D volume scanning. This resulted in the maximal 3D diameter X and the follicular volume (V). Each follicle was given a code, and its volume was recorded on the ultrasound scans as well as its content. Another physician corroborated all ultrasonographic measurements and the erroneous identification of follicles underwent manual correction. Then the volume of the irregular follicle(V) was subsequently used as the equivalent sphere volume (the same spherical volume as the follicle volume) and the equivalent sphere diameter (d v ) was derived from the formula. Finally, the ratio of the maximum follicle 3D diameter(X) to the equivalent sphere diameter(d v ) was compared to obtain the follicular sphericity (Fig.  2 ) [ 23 , 24 ]. The mean value of the sphericity of all follicles then represented the follicular sphericity of one patient. The relationship between follicular sphericities and their locations is shown from the ovarian cortex to the medullary region in Fig.  3 . The calculations were as follows: 1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${d_v} = 2 \times \root 3 \of {\frac{3}{4}{v \mathord{\left/{\vphantom {v \pi }} \right.\kern-\nulldelimiterspace} \pi }} $$\end{document} and 1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Sphericity = \frac{{{d_v}}}{{\max imal\,3D\,diamter\,X}}$$\end{document} Fig. 2 Measurement of follicular sphericity on the HCG trigger day by 3D-US monitoring *The follicular volume ( V ) and the diameter ( d v ) of the equivalent sphere for this volume were obtained by 3D-US follicle monitoring on the HCG trigger day. After receiving the minimal bounding sphere by the maximal 3D diameter X, the follicular sphericity was achieved by comparing its similarity to the equivalent sphere ( d v / X ) Measurement of follicular sphericity on the HCG trigger day by 3D-US monitoring *The follicular volume ( V ) and the diameter ( d v ) of the equivalent sphere for this volume were obtained by 3D-US follicle monitoring on the HCG trigger day. After receiving the minimal bounding sphere by the maximal 3D diameter X, the follicular sphericity was achieved by comparing its similarity to the equivalent sphere ( d v / X ) Fig. 3 Relationship between follicular sphericity and follicle locations within the ovary in 3D-US images A1 : follicular sphericity in patients with normal ovarian reserve; A2 : follicular sphericity in patients with poor ovarian reserve Relationship between follicular sphericity and follicle locations within the ovary in 3D-US images A1 : follicular sphericity in patients with normal ovarian reserve; A2 : follicular sphericity in patients with poor ovarian reserve After the HCG trigger, transvaginal oocyte retrieval was performed at 36 h, and embryo transfer was conducted after oocyte retrieval. Most patients underwent transfer of two cleavage-stage embryos unless the patient received a single blastocyst or only had one available embryo to transfer, following national guidelines [ 25 ]. Luteal-phase support was initiated immediately after oocyte retrieval with combined vaginal and oral progesterone. Surplus available embryos were frozen for later transfer in subsequent frozen embryo transfer (FET) cycles. The primary outcome indicators reflected ovarian responsiveness as well as IVF success. The rate of MII oocytes was calculated as the number of MII oocytes per retrieved oocytes; the rate of normal fertilization was calculated as the number of normally fertilized oocytes per MII oocyte; and the rate of Day 3−8cell (D3−8 C) was defined as the proportion of embryos with eight cells on Day 3 relative to the total number of cleavage embryos with two pronuclei (2PN). The follicular output rate (FORT, a potentially outstanding tool used to evaluate follicle responsiveness [ 26 ] was calculated as the ratio between preovulatory follicle count (PFC) and 100/AFC on the HCG day; we only considered follicles ≥ 14 mm for the calculation of FORT. The ovarian sensitivity index (OSI, oocytes per 1000 IU) as a measure of ovarian response [ 27 ] was calculated by the number of retrieved oocytes per 1000/Gn total dose on the HCG day. Ovarian reserve was not considered as either an inclusion or exclusion criterion. A general linear model (GLM) analysis was executed to adjust ovarian reserve (AMH, AFC) and relevant Gn treatment so as to control for bias. Data that followed a normal distribution are presented as mean (standard deviation, SD) and the median was considered descriptive for data that were not normally distributed. Analysis of variance (T-test) was applied to determine the differences in group comparisons, and classified variables were compared using the Chi-squared test. Multivariate linear regression was implemented to explore the influence of follicular sphericity. Furthermore, the ROC curve analysis was constructed using MedCalc statistical software (MedCalc version 20.0.26, Mariakerke, Belgium) and the Youden index was used to determine the cut-off value of the follicular sphericity with respect to the MII oocyte rate. We performed a repeated-measures ANOVA test to uncover correlations between ovarian volume in the groups. Partial correlation analysis and the GLM were used to compare the number of retrieved oocytes, rate of oocytes progressing to MII, FORT, OSI, and D3−8c rate between groups after controlling for confounders. P -values < 0.05 were considered to indicate statistical significance. We conducted all analyses using Stata version 15.1 (StataCorp LLC, College Station, TX, USA) and RStudio Team 2015 (RStudio, Inc., Boston, MA, USA).

This was a prospective, single-cohort observational study, and congruent with our inclusion and exclusion criteria, we identified a total of 248 patients as eligible for participation in this analysis. The follicular sphericity of the left and right ovaries ranged from 0.29 to 0.89 and the mean value for follicular sphericity for patients(i.e., mean value of left and right ovarian follicular sphericity) was 0.70 ± 0.04, and the medians and quartiles for sphericity were 0.68 and 0.72, respectively (the distribution is shown in Supplement Fig.  2 ). We exploited multivariate linear regression based on the follicular sphericity with regard to patient characteristics. The average follicular sphericity of all follicles on HCG was calculated as the follicular sphericity in patients. After considering confounders (such as age, BMI, AMH, FSH, luteinizing hormone (LH), estradiol(E 2 ), AFC, infertility factors and fertilization), the AFC acted as an independent factor influencing follicular sphericity for patients. We discerned that follicular sphericity significantly decreased with the increase in AFC (adjusted β=−0.215, P  value = 0.020)(Table  1 ). Table 1 Multivariable linear regression of factors influencing follicular sphericity Unadjusted β coefficients adjusted β P Age 0.001 0.121 0.108 BMI, kg/m 2 0.001 0.09 0.238 Serum AMH, ng/mL -0.001 -0.092 0.354 Base serum P, IU/L 0.014 0.075 0.254 Base serum FSH, IU/L 0.001 0.059 0.424 Base serum LH, IU/L -0.001 -0.055 0.461 Base serum E 2 , pg/ml 0.001 -0.034 0.599 Infertility factors 0.001 0.007 0.923 Fertilization 0.001 0.093 0.456 Number of AFC 0.001 -0.215 0.020 *The infertility factors included male factor, ovulatory dysfunction, tubal factor, diminished ovarian reserve and unexplained factor; fertilization included ICSI and IVF *Abbreviations: follicle stimulating hormone(FSH), antral follicle counts(AFC), anti-Müllerianhormone(AMH), body mass index(BMI), Luteinizing hormone (LH), estradiol(E2) Multivariable linear regression of factors influencing follicular sphericity *The infertility factors included male factor, ovulatory dysfunction, tubal factor, diminished ovarian reserve and unexplained factor; fertilization included ICSI and IVF *Abbreviations: follicle stimulating hormone(FSH), antral follicle counts(AFC), anti-Müllerianhormone(AMH), body mass index(BMI), Luteinizing hormone (LH), estradiol(E2) Figure 4 shows the ROC curve analysis of follicular sphericity with respect to MII oocytes, and we calculated the cut-off values with the Youden index. The cut-off value for follicular sphericity was 0.716 (sensitivity = 0.723, specificity = 0.463), showing from ROC results that follicular sphericity (AUC = 0.608 [0.530–0.686]) was a predictor of oocyte rate to MII. In order to verify whether patients with lower follicular sphericity could increase the number of retrieval oocytes, patients were assigned by cut-off values into two groups: follicular sphericity 82.14%). Epidemiologic, clinical, and biologic characteristics of the study population are summarized in Table  2 . Patients differed in ovarian reserve and the dose of their Gn treatment, AMH levels, and their AFC—with the values in the high sphericity group lower than in the low sphericity group (all P  values < 0.05); and they received a higher Gn starting dose in the low sphericity group(all P  values < 0.01). After the patients were allocated into poor ovarian reserve group(AFC < 5) and normal ovarian reserve group(AFC ≥ 5), we noted a higher AFC number and lower Gn starting dose in the low sphericity group than in the high sphericity group (12.15 ± 4.57 vs. 9.74 ± 4.75; 176.14 ± 42.35 vs. 203.88 ± 52.78, respectively; P  value < 0.01) in patients with normal ovarian reserve. In the repeated-measures ANOVA test, we observed a significant difference from the Gn day start until the HCG day in changes of ovarian volume between groups(F = 6.88, P  = 0.009). There was no difference in patients with poor ovarian reserve (Supplement Table  1 ). Fig. 4 Follicular sphericity in the prediction of MII oocyte by receiver operator characteristic (ROC) curve *the red arrow depicts the cut-off value Follicular sphericity in the prediction of MII oocyte by receiver operator characteristic (ROC) curve *the red arrow depicts the cut-off value Table 2 Baseline characteristics of patients undergoing IVF treatment Low sphericity ( n  = 171) High sphericity ( n  = 77) P  value Age 31.07 ± 3.82 32.48 ± 4.34 0.010 BMI, kg/m 2 22.37 ± 3.09 22.67 ± 2.97 0.473 Serum AMH, ng/mL 4.4 ± 2.76 3.34 ± 3.05 0.007 Base serum P, IU/L 0.25 ± 0.2 0.28 ± 0.13 0.327 Base serum FSH, IU/L 5.97 ± 1.57 6.4 ± 1.7 0.055 Base serum LH, IU/L 3.47 ± 1.84 3.24 ± 1.5 0.331 Base serum E 2 , pg/ml 29.14 ± 13.92 29.02 ± 14.4 0.953 Number of AFC 10.82 ± 5.19 8.38 ± 4.81 0.001 Starting dose of Gn 187.72 ± 50.87 214.29 ± 51.89 < 0.001 Days of Gn administration 8.72 ± 5.7 7.92 ± 1.29 0.226 Total dose of Gn 1584.72 ± 654.08 1700.65 ± 506.77 0.169 Infertility factors  Male factor 17.55% (30) 12.99% (10) 0.352  Ovulatory dysfunction 11.70% (20) 9.09% (7)  Tubal factor 66.67% (114) 68.83% (53)  Diminished ovarian reserve 1.75%(3) 5.19%(4)  Unexplained factor 2.33%(4) 3.90%(3) Fertilization  ICSI 26.90% (46) 27.27% (21) 0.882  IVF 73.10% (125) 72.73% (56)  Basal ovarian volume(cm 3 ) 4.70 ± 2.64 3.84 ± 1.98 0.015  Ovarian volume on HCG day(cm 3 ) 22.11 ± 11.12 15.54 ± 8.09 0.001 *Abbreviations: body mass index(BMI), anti-Müllerian hormone(AMH), follicle-stimulating hormone(FSH), antral follicle count(AFC), in vitro fertilization(IVF), intracytoplasmic sperm injection(ICSI), Luteinizing hormone (LH), estradiol (E 2 ) Baseline characteristics of patients undergoing IVF treatment *Abbreviations: body mass index(BMI), anti-Müllerian hormone(AMH), follicle-stimulating hormone(FSH), antral follicle count(AFC), in vitro fertilization(IVF), intracytoplasmic sperm injection(ICSI), Luteinizing hormone (LH), estradiol (E 2 ) Adopting the Kolmogorov-Smirnov test, we found the parameters of interest to be non-normally distributed. Applying the Wilcoxon signed-rank test, we observed that the number of retrieved oocytes, MII oocyte rate, FORT, OSI, D3−8c rate and implantation rate were significantly lower in the high sphericity group relative to the low sphericity group (all P  values < 0.05) in patients with normal ovarian reserve. Our partial correlation analysis after adjusting for age, AMH, AFC, and Gn treatment revealed that high sphericity was negatively correlated with the embryonic results; with the greater number of retrieved oocytes, MII oocyte rate, FORT, OSI, D3−8c rate and implantation rate with the lower the sphericity. Fertilization and pregnancy rates did not differ among groups (Table  3 ). However, we discerned that no such difference in patients with low ovarian reserve. Table 3 Ovarian responsiveness and embryonic and clinical outcomes between groups Patients with poor ovarian reserve Patients with normal ovarian reserve Low sphericity ( n  = 28) High sphericity ( n  = 19) P Low sphericity ( n  = 143) High sphericity ( n  = 58) P Retrieval Oocytes 5.00 (3.75, 7.25) 5.50 (4.25, 6.75) 0.631 15.00 (9.00,20.00) 10.00 (6.00,13.00) < 0.001 MII rate (%) 85.00 (75.00,94.74) 85.42 (72.32,98.91) 0.631 89.71 (75.43,96.44) 84.21 (69.81,92.00) 0.024 FORT (%) 75.02 (61.16,91.25) 63.33 (44.64,88.33) 0.343 78.34 (52.94,89.10) 66.67 (42.86,85.18) 0.007 OSI 2.67 (1.71,4.26) 2.84 (2.32,4.44) 0.312 10.37 (6.15,16.00) 5.71 (3.47,8.85) < 0.001 Fertilization rate (%) 67.21 (56.06,83.09) 62.73 (55.56,87.09) 0.865 68.55 (59.80,89.11) 66.45 (58.12,86.31) 0.359 D3-8c rate (%) 33.12 (11.11,50.00) 32.46 (15.38,50.00) 1.000 48.57 (27.14,71.83) 36.67 (12.34,56.87) 0.004 Implantation rate (%) 9/41 (21.95) 3/18 (16.67) 0.845 100/198 (50.50) 32/87 (36.78) 0.035 Pregnancy rate (%) 7/26 (26.92) 3/12 (25.00) 0.967 73/143 (51.05) 26/58 (44.83) 0.441 Correlation coefficient β was conducted by partial correlation analysis after adjusting Age, AMH, AFC, and Gn treatment *Abbreviations: follicular output rate(FORT), ovarian sensitivity index(OSI), Day 3-8cell(D3-8 C) Ovarian responsiveness and embryonic and clinical outcomes between groups Correlation coefficient β was conducted by partial correlation analysis after adjusting Age, AMH, AFC, and Gn treatment *Abbreviations: follicular output rate(FORT), ovarian sensitivity index(OSI), Day 3-8cell(D3-8 C) In the general linear model for univariate (GLM-univariate) analysis and controlling the confounders of AFC, Gn treatment and ovarian volume changes of patients with normal ovarian reserve, we noted that the number of retrieved oocytes, rate for MII oocytes, FORT, OSI, and D3−8c rate were significantly lower in the high sphericity group than in the low sphericity group(all P  values < 0.01) (the associations between sphericity and oocyte or embryo quality are shown in Table  4 ). Compared with women in the high sphericity group, women in the low sphericity group had a higher oocyte retrieval rate, MII oocyte rate, FORT, OSI and D3−8c rate(adjusted β = 5.46, 5.15, 1.08, 2.39, 1.17, respectively; all P  value < 0.05). Ultrasound images of patients with more follicles showed that the follicles became more irregular as they developed forward to the interior of the ovary (Fig.  4 -A1), while ultrasound images of patients with fewer follicles showed that the follicles became rounder or migrated slightly from the cortical to medullary regions of the ovary (Fig.  4 -A2). We performed ROC curve analysis to evaluate the significance of AFC on MII oocytes, and the AUC of the AFC was 0.712 (95% CI, 0.655–0.779, P  value < 0.001), improving to 0.773 (95% CI, 0.709–0.836, P  value  0.75) (Supplementary Fig.  3 ). Table 4 General linear model predicting the effect of adjusted sphericity on ovarian responsiveness and embryonic and clinical results of patients with normal ovarian reserve Oocytes retrieved MII rate FORT OSI D3-8 C rate Adjusted β (95% CI) P  value Adjusted β (95% CI) P  value Adjusted β (95% CI) P  value Adjusted β (95% CI) P  value Adjusted β (95% CI) P  value Sphericity group   High sphericity Ref Ref Ref Ref Ref   Low sphericity 5.46(2.74–7.51) 0.013* 5.15 (4.01–4.38) 0.007* 1.08 (1.01–1.15) 0.006* 2.39(1.71–5.30) 0.010* 1.17 (1.05–1.23) 0.008* AFC 1.99(1.60–2.48) < 0.001* 1.72(1.43–2.06) < 0.001* 1.04(1.03–1.05) < 0.001* 1.98(1.61–2.45) < 0.001* 0.99(0.98–1.01) 0.505 Starting dose of Gn 0.99 (0.97–1.01) 0.319 1.00 (0.99–1.00) 0.721 1.00 (0.99–1.00) 0.709 1 (0.98–1.02) 0.896 1.00 (0.99–1.00) 0.935 Volume changes of the ovarian volume 3.21 (2.31–4.46) < 0.001* 1.09 (1.06–1.11) < 0.001* 1.03 (1.02–1.05) < 0.001* 1.97 (1.55–2.51) < 0.001* 1.01 (0.99–1.02) 0.343 All analyses were conducted using generalized linear models. Data are presented as predicted marginal proportions (95% CI) adjusted for AFC, Gn starting dose * P -Value < 0.05 for comparisons with the reference category *Abbreviations: AFC, antral follicle count General linear model predicting the effect of adjusted sphericity on ovarian responsiveness and embryonic and clinical results of patients with normal ovarian reserve All analyses were conducted using generalized linear models. Data are presented as predicted marginal proportions (95% CI) adjusted for AFC, Gn starting dose * P -Value < 0.05 for comparisons with the reference category *Abbreviations: AFC, antral follicle count

The principal purpose of assisted reproductive technology (ART) treatments is to achieve pregnancy. However, as oocyte retrieval is considered to be an important prognostic variable [ 1 ], treatment protocols aim to optimize the ultimate outcome. Controlled ovarian stimulation (COS) generates follicles of different sizes, and this process is monitored by transvaginal sonography; human chorionic gonadotropin (HCG) is then usually administered at the end of COS to induce final follicular maturation [ 2 ]. Fertility specialists determine whether the follicle size, number, and serum hormone concentrations meet trigger requirements so as to prevent the retrieval of immature oocytes. Oocyte maturity is known to be related to follicle size [ 3 ], and the most widely applied protocol initiates the trigger when several follicles reach a diameter of > 17–18 mm [ 4 ]. Nevertheless, several investigators have recently questioned the accuracy of the two-dimensional (2D) measuring approach mentioned above—especially in COS cycles— since the shape of follicles is irregular [ 5 , 6 ]. The measurement of follicle size is thus often inaccurate, even in ovaries containing small or few follicles, as the evaluation is conducted on follicles considered to be 2D structures that are in fact three-dimensional (3D) structures. Follicles rarely possess a perfectly spherical shape and are frequently elliptical, especially with follicular overcrowding in a hyperstimulated ovary [ 7 ]. The mean diameter measured manually by 2D ultrasound (2D-US) neglects the 3D information of the follicle shape, and may not reflect the actual diameter of the follicle [ 8 , 9 ]. It has been suggested that the underestimation of 2D results was due to the inherent difficulty in precisely placing the US probe simultaneously on the perpendicular maximum of the x and y follicle diameters, leading to measurement errors that (in theory) were normally distributed [ 10 ]. The current literature entailing the precision of 3D versus 2D ultrasound in IVF follicular assessment indicates that 3D ultrasound provides superior efficacy and accuracy in measuring follicular dimensions and volume compared to real-time 2D ultrasound [ 10 ]. A notable advantage of 3D ultrasound is evident when imaging ovaries containing over 20 follicles [ 11 ]. Several articles were subsequently published in which 2D-US and 3D-US were compared for follicular monitoring. Although there were some variations in the measurement protocols, there were differences between follicular monitoring using 2D-US and the SonoAVC follicle software in terms of the higher numbers of fertilized oocytes in the 3D group [ 12 ], that did not translate into significantly higher numbers of mature oocytes retrieved or pregnancy rates [ 13 , 14 ]. 3D ultrasound (3D-US), in addition, has been used in the diagnosis of patients with polycystic ovary syndrome (PCOS), and studies have shown that the ovaries of patients with PCOS tend to be more spherical [ 15 , 16 ]. However, studies have suggested that stimulated follicles are often not perfectly spherical, and that the automatic method is more accurate and reproducible, providing more morphologic information on follicles that are not accessible with 2D-US [ 17 , 18 ]. With 3D-US, it is possible to obtain more information regarding follicular morphology, including the sphericity of the follicles. However, the impact of follicle sphericity on oocyte quality remains arcane considering the mechanical role of follicles in the ovary. We herein explored novel follicular parameters using 3D-US, and compared the follicular sphericity on the day of HCG administration as measured with 3D-US during gonadotropin-releasing hormone antagonist (GnRH-ant) stimulation cycles. We additionally investigated the effect of sphericity on ovarian responsiveness and oocyte and embryo quality.

Many investigators have over the past 15 years attempted to predict oocyte retrieval during stimulation. A significant number of them evaluated the efficacy of biological markers, including hormone levels [ 28 – 31 ], alongside factors such as gonadotropin treatment, age, BMI, and infertility type. Research on predicting retrieved oocytes in IVF is scant and often limited to small sample size, poor model accuracy, and prediction efficacy due to inadequate objective parameters [ 32 , 33 ]. Ultrasonic follicular assessment is essential in IVF and ovarian stimulation. 2D ultrasound is typically implemented to measure follicle diameter and quantity to monitor progression, predict oocyte maturation, and optimize the timing of the HCG trigger [ 17 ]. However, application of 2D-US assumes that follicles are standardized in shape, and this is often inaccurate as they can be asymmetrical [ 6 ]. 3D-US has recently emerged as a superior alternative for precise follicular monitoring, with SonoAVC as a notable example. Previous studies have revealed a discrepancy between 3D-US and 2D-US measured follicular diameter, with the mean difference ranging from 1.02 mm to 0.9 mm [ 34 , 35 ]. We postulate that the mathematical principles of 3D-US calculations may be able to explain these differences. Volume-based diameter is the relaxed sphere diameter of a perfect sphere with the same volume as the follicle and the mean of three maximal diameters of an irregular follicle is always greater than the diameter of a perfect sphere of the same volume. This discrepancy tended to increase as follicle size increased, and thus 3D-US correlated strongly with 2D metrics and provided more parameters with respect to follicular development [ 9 , 36 ]. This study suggests that based on 3D transvaginal ultrasound, follicular sphericity could predict ovarian responsiveness during routine IVF monitoring in patients with normal ovarian reserve. As AFC acts as an independent influence on follicular sphericity, it remains predictive of the number of retrieved oocytes, MII oocyte rate, FORT, OSI, D3−8c rate, and implantation rate when controlling for AFC as a confounding factor. The follicular sphericity may serve as an important measure in managing ovarian stimulation, and reflects the development of follicles in COS, thus providing insights for adjusting gonadotropin drug regimens. However, the predictive effectiveness of follicular sphericity was inadequate in our patients with poor ovarian reserve. In our study, the ROC curve for follicular sphericity yielded an AUC of 0.608 for predicting MII oocytes, which only reflects the morphology of follicle development and does not involve multifactorial regulation beyond the parameters measured(e.g., molecular biomarkers and follicular fluid microenvironment). While the AUC for follicular sphericity is limited to standalone clinical applications, our findings suggested that follicular sphericity may still be valuable as a supplementary parameter in multimodal prediction models. The results of the present study indicated that AFC, an ultrasound biomarker of follicle number, has been widely used in clinical practice. It is regarded as the most reliable and accurate marker of ovarian reserve [ 37 ]. AFC refers to the cumulative number of follicles observed via ultrasonography in both ovaries during the initial phase of the menstrual cycle, specifically days 2–4. Antral follicles are follicles that exhibit a maximal mean diameter of 2–10 mm when measured in the two-dimensional plane [ 38 ]. AFC is widely considered to be the most dependable approach for evaluating the ovarian response to ovarian stimulation. Nevertheless, the outcome relies on the operator’s expertise and the precision of the ultrasonogram [ 39 , 40 ]. The greatest advantage of AFC over AMH is the ability of the observer to evaluate many other important aspects of the ovaries when assessing the functional ovarian reserve by AFC, such as their position and the presence of endometrioma or other ovarian lesions—as well as gaining important information regarding the Fallopian tubes (e.g. presence of hydrosalpinx) and the uterus (e.g. presence of endometrial polyps or submucous leiomyomas) [ 5 ]. Although follicle count is often evaluated by two-dimensional (2D)-US, three-dimensional (3D)-US might also be used [ 15 ]. 3D-US imaging has the advantage of a shorter examination time, as it enables storage of acquired data for offline analysis, and superior interobserver reliability [ 41 ]. In order to maximize reliability, the follicle count should be performed only by a competent observer; however, there is no specific training or certification. Furthermore, there is no consensus as to how many scans a person should perform before being considered technically competent [ 5 ]. Although influenced by AFC and ovarian volume, the follicular sphericity, as a new ultrasonographic parameter, can also be used as a morphologic parameter to show follicular development in the ovary. The principle of sphericity for the assessment of ovarian responsiveness and oocyte quality has not been fully investigated and validated in the existing literature. Therefore, in order to explore the potential significance of sphericity in ART, its possible underlying mechanism of the ovarian internal structure and the ability of follicles to expand will be discussed in the following points. Estrogen is likely to play a role in arterial distensibility, but the available evidence suggests that its effect depends on the baseline estrogen level and magnitude of the estrogen concentration changes [ 42 ]. The process of COS is accompanied by changes in serum hormone levels, of which changes in E 2 levels exert a significant impact on the overall COS. Available evidence suggests that COS leads to a significant increase in E 2 levels, which may affect blood pressure through multiple mechanisms. Studies have confirmed that a hyperestrogenic state may trigger vascular endothelial dysfunction; physiological levels of E 2 maintain vasodilation by promoting the release of nitric oxide (NO) and prostacyclin (PGI2), but excessively high E 2 reduces NO bioavailability and leads to vasoconstriction [ 43 ]. Second, E 2 activates the renin-angiotensin-aldosterone system (RAAS), which increases angiotensinogen synthesis, promotes sodium retention and blood volume rise, and thus elevates blood pressure [ 44 ]. In addition, estrogen enhances renal reabsorption of sodium through antidiuretic hormone (ADH), exacerbating fluid retention [ 6 ]. It has also been shown that ART-assisted pregnancy patients experience changes in E 2 from baseline to up to tenfold concentrations during pregnancy, which results in a one-third reduction in carotid artery elasticity [ 42 , 45 ] and their increased risk of gestational hypertension and preeclampsia may be associated with COS-induced supraphysiologic fluctuations in E 2 levels [ 42 , 46 ]. To elucidate ovarian and follicular structures, researchers currently focus on scaffolding, matrix proteins, and systems that deepen our understanding of the growth prerequisites of follicles. The spherical shape of the ovarian follicle is maintained through essential cell-cell and cell-matrix interactions within the surrounding stromal tissue, enabling follicles to effectively complete the maturation process [ 47 , 48 ]. Cells react to mechanical stimuli as exemplified by substrate stiffness by modifying properties such as stiffness, motility, adhesion, and contractility [ 49 ]; and the cytoskeleton’s organization regulates these mechanical traits [ 50 ]. Curvature generation by proteins is influenced by the membrane’s mechanical characteristics, including tension and bending rigidity. Cytoskeletal forces significantly affect membrane deformation by adjusting effective tension through membrane-to-cortex interactions [ 51 , 52 ]. Thus, cytoskeletal tension impacts the ability of curvature-inducing proteins to generate curvature. Studies show that reduced cortical tension may increase curvature propensity, while heightened tension in a stiff extracellular matrix (ECM) may remain within a saturated region, minimally affecting the formation of highly curved structures [ 53 ]. Emerging evidence suggests that the ECM is vital for follicle development, either suppressing or enhancing growth; and that this mechanical environment is controlled in turn by hormones or growth factors. Studies show that increased stiffness hinders follicle growth, disturbs the androgen-to-estrogen ratio, and produces gametes with reduced meiotic potential [ 54 – 56 ]. Primordial follicles reside in the collagen-rich ovarian cortex that provides the rigid environment essential for follicular structure and survival [ 57 , 58 ]. The stiffness of the cortical ECM restricts follicular expansion and oocyte maturation, preserving the follicles in a dormant state [ 59 , 60 ]. Scientists have proposed that ovarian biomechanics influence primordial follicle activation, with the dense cortex maintaining quiescence while the peri-medullar region fosters growth. Considering the internal cytoarchitecture of the ovary, researchers ascertained that large follicles were the mechanically dominant structures in the ovary that exhibit substantial mechanical variations, particularly within the stiffer regions of the ovary [ 61 ]. However, the biomechanical control of follicular development in the ovary remains hypothetical. During folliculogenesis, activated follicles migrate from the ECM of the dense outer cortex to the inner, softer peri-medullary zone where they grow in size and facilitate oocyte maturation until ovulation at the ovarian surface. We obtained the same configurations in our 3D-US images of a human female ovary. These findings further support the hypothesis that mechanical heterogeneity between the cortex and medulla in the ovary plays an integral role in fertility [ 61 , 62 ]. Based on the above description, a category of good-quality follicles with a higher volume of follicular fluid will produce lower sphericity and migration from the cortex to the medulla of the ovary. Our low-sphericity group demonstrated a greater tendency for positive clinical outcomes, but possessed a relatively small sample size, and thus, additional large-sample trials are needed to corroborate this result. Notably, we have only observed the predictive efficacy of follicular sphericity on ovarian responsiveness and embryo quality in patients with normal ovarian reserve (AFC ≥ 5). Follicular sphericity did not correlate with any clinical outcomes in patients with diminished ovarian reserve. The ROC curve for follicular sphericity yielded an AUC of 0.608 for predicting MII oocytes, only reflecting the morphology of follicle development and not involving multifactorial regulation beyond the parameters measured—limiting standalone clinical application. Follicular sphericity as a real-time (compared to AFC) and more realistic (compared to 2D follicular diameter) assessment could be applied to monitor daily follicular development and might be used as a powerful tool for Gn modification. But considering the limitations of sample size and lack of external validation, a large sample size, multicenter cohort study is warranted to further validate our results. Despite variability in ovarian reserve among patients, the bilateral ovaries consistently exhibit different dynamics in volume based on low- and high-sphericity classifications. This physiologic insight underscores the reliability of follicular sphericity as a parameter that can be adopted for assessing ovarian responsiveness and also indicates its potential in predicting ovarian response, embryonic development, and clinical outcomes. Our 2D to 3D image algorithm can be integrated with EMRS, enabling the creation of a fully automated, multimodal dataset repository. We posit that this development will facilitate the future clinical application of this technology.

This study was the first-ever to show that based on 3D-US algorithms, follicular sphericity could be used to reliably predict ovarian responsiveness during routine IVF monitoring in patients with normal ovarian reserve. We posit that follicular sphericity will serve as an important measure in managing ovarian stimulation, providing insights for adjusting gonadotropin drug regimens.

Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1