An electronic search of the radiology information system from July 2018 to June 2019 was performed in two hospitals, and 3473 inpatients underwent pelvic MRI. Then, the electronic medical records were reviewed. In all, 308 were pregnant women, 70 cases of which were excluded, and 238 cases were enrolled in the final analysis. Controls were first matched to cases based on: hospitalization period (same time frame to minimize temporal bias) and Age ($\pm$1 year to ensure close demographic alignment). If multiple controls met these criteria for a single case, we further selected the individual with the most similar number of vaginal deliveries (though this was not a strict matching factor). 238 controls were enrolled in the final analysis. (Supplementary Fig. 1).

Exclusion criteria of pregnant women were as follows: (1) a history of pelvic organ prolapse (POP) or urinary/fecal incontinence; (2) any mass or cyst in the cervix or the cervical canal with a diameter $\ge$2 cm; (3) any mass or cyst on the vaginal wall or in the vaginal lumen with a diameter $\ge$2 cm; (4) any mass or cyst in the pelvis with a diameter $\ge$6 cm, such as fibroids or ovarian cysts; (5) the rectum without emptying having a maximum diameter $\ge$3 cm or the bladder without emptying having a maximum diameter $\ge$6.5 cm on any sagittal images; (6) poor visibility of the anatomic landmarks; (7) a history of pelvic fracture; (8) ectopic pregnancy; (9) multiple pregnancies; or (10) uterine incarceration.

Controls were women with a normal uterus and no POP indicated by the gynecological examination record [18]. Exclusion criteria of controls were as follows: (1) any exclusion criterion for pregnant cases; (2) presence of POP on gynecological examination; (3) enlarged uterus with a maximum diameter $\ge$7 cm, such as multiple uterine fibroids or adenomyosis; and (4) within 12 months postpartum.

In our two hospitals, pelvic MRIs were performed using a 1.5 Tesla magnet (Sigma, General Electric Medical System, Milwaukee, WI, USA) with patients in a lying position. The MRI measurements were performed on the (near-) midline sagittal plane by an experienced gynecologist-obstetrician and radiologist, using the Web-viewer (Version 1.0.0.53449, Greenlander Information Technology, Hangzhou, Zhejiang, China). Discrepancies were resolved through discussion. The sequences reviewed in this study were fat-saturated T2-weighted fast recovery fast spin-echo, T2-weighted single-shot fast spin-echo, or two-dimensional (2D) fast imaging employing the steady-state acquisition technique.

The pubococcygeal line (PCL), the recommended MRI reference line for diagnosing POP, extends from the pubic symphysis’s inferior aspect to the last coccygeal joint [19]. Perpendicular distances from the PCL to the bladder neck (BN), the cervix’s most distal edge (C), and the posterior fornix’s apex (PF) were measured (Fig. 1). Organ-specific reference points above the PCL had positive values (+), while those below had negative values (–).

Fig. 1.

Sagittal MRI-based measurement of pelvic organ positions relative to the pubococcygeal line (PCL). PCL is the red line from pubic symphysis to the last coccygeal joint. Perpendicular distances from the PCL to anatomical points BN, C, and PF are measured (denoted by short red lines) in non-pregnant women (A) and pregnant women with cephalic presentation (B). BN, bladder neck; C, most distal cervical edge; MRI, magnetic resonance imaging; PF, apex of the posterior fornix; PS, pubic symphysis.

An additional line drawn from the pubic symphysis’ superior margins to the sacral promontory defines the pelvic inlet plane for pregnant women [20]. The depth of the fetus and its appendages in the pelvis was defined as the greatest distance from the pelvic inlet plane to the fetal presentation or the inferior part of the placenta or amniotic sac (Fig. 2). All measurements were performed in millimeters.

Fig. 2.

Engagement depth of fetal/placental structures relative to pelvic inlet plane. (A) Cephalic presentation with shallow engagement (Depth = 22 mm). (B) Cephalic presentation with deep engagement (Depth = 65 mm; amniotic sac as the most inferior part entering the pelvis). (C) Placenta previa (Depth = 40 mm; placenta as the most inferior part entering the pelvis). (D) Breech presentation (Depth = 60 mm; amniotic sac as the most inferior part entering the pelvis). Blue solid line: Pelvic inlet plane, drawn from the superior margin of the pubic symphysis to the sacral promontory. Blue dashed line: Parallel to the pelvic inlet plane, passing through the lowest point of the fetus or its associated structures. Depth: Perpendicular distance between the two parallel lines, quantifying the extent of fetal/placental descent into the pelvis. Red solid line: Pubococcygeal line (PCL).

Descriptive variables, including baseline data and MRI measurements, were reported as percentages (n (%)), medians (interquartile range), and means $\pm$ standard deviation (SD). The Kolmogorov-Smirnov test was employed to assess the normality of the data distribution. Pelvic organ positions in the pregnant group were classified as “inferior” or “superior” using thresholds based on the mean $\pm$ 1.96 SD from the non-pregnant group. Univariate logistic regression analysis explored the relationships between pelvic organ positional variations and factors such as age, parity, vaginal delivery history, and obstetric factors including gestational age at MRI, body mass index (BMI) at MRI, depth of fetus and its appendages in the pelvis, fetal presentation, and placental location. Gestational age and the other factors with a p-value 0.05 in univariate analysis were included in multivariate analysis. Before multivariate analysis, interactions between factors were evaluated using correlation analysis. The Hosmer-Lemeshow test was used to assess the goodness-of-fit and the Nagelkerke R² values were also calculated to illustrate the explanatory power of the models. Associations were quantified using odds ratios (ORs) and 95% confidence intervals (CIs). Statistical analyses were performed using IBM SPSS Statistics 23.0 (IBM Corp., Armonk, NY, USA), with p-values 0.05 considered statistically significant.

The average gestational age of the 238 pregnant women was 32.3 $\pm$ 5.8 weeks, ranging from 12 to 41 weeks. Of these, 182 (76.5%) was the cephalic presentation. The primary indication for pelvic MRI was the suspicion of an abnormal placenta, which accounted for 192 (80.7%) of the cases. Among these, 122 (51.3%) were ultimately diagnosed with a low-lying placenta or placenta previa. The average depth of the fetus and its appendages in the pelvis was 32.6 $\pm$ 16.4 mm (Supplementary Table 1). The primary clinical indications for MRI in the non-pregnant control group included adnexal mass evaluation (47.9%) and uterine pathology assessment (35.7%).

There was no significant difference in age at the time of MRI between the two groups (32.2 $\pm$ 5.1 years for the pregnant group vs. 32.2 $\pm$ 5.3 years for the non-pregnant group, p = 0.909) or in the number of vaginal deliveries (p = 0.211). However, the pregnant group had higher BMI levels, parity, and frequency of previous caesarean deliveries than the non-pregnant group (p 0.001 for all). In comparison with the non-pregnant group, the pregnant group showed inferior BN (19.2 $\pm$ 6.1 mm vs. 23.3 $\pm$ 4.3 mm, p 0.001) and C positions (19.1 $\pm$ 7.8 mm vs. 21.6 $\pm$ 5.3 mm, p 0.001), but a superior PF position (42.3 $\pm$ 10.5 mm vs. 31.7 $\pm$ 8.1 mm, p 0.001) (Table 1).

Among the 238 pregnant women studied, 53 (22.3%) were identified with an “inferior” BN position, indicated by values lower than 14.87 mm. Additionally, 39 (16.4%) exhibited an “inferior” C position with values lower than 11.21 mm, and only one (0.42%) had an “inferior” PF position, with values below 15.82 mm. In contrast, 71 (29.8%) cases showed a “superior” PF position with values exceeding 47.58 mm.

Univariate analysis indicated that greater gestational age, deeper depth of the fetus and its appendages in the pelvis, and an increased number of vaginal deliveries were potentially associated with an “inferior” BN position (Supplementary Table 2). Multivariate logistic regression (Fig. 3A) showed that with each additional week of gestational age, the odds of an “inferior” BN position increase by 12% (p = 0.004). And an “inferior” BN position was found in 45.3% of women with entering pelvis depth $\ge$40 mm compared to 9.4% in those 20 mm, an odds ratio of 5.04 (95% CI 1.59–16.01, p = 0.006). Moreover, with each additional vaginal delivery, the odds of an “inferior” BN position increased by 2.95 (95% CI 1.41–6.20, p = 0.004).

Fig. 3.

Multivariable logistic regression analysis of pelvic organ positions. In the multivariate logistic regression model for bladder neck (A), the Hosmer-Lemeshow test yielded $\chi$² (8) = 4.124, p = 0.846, and the Nagelkerke R² was 0.206. In the multivariate logistic regression model for cervix (B), the Hosmer-Lemeshow test yielded $\chi$² (8) = 7.498, p = 0.484, and the Nagelkerke R² was 0.245. In the multivariate logistic regression model for posterior fornix (C), the Hosmer-Lemeshow test yielded $\chi$² (8) = 6.900, p = 0.547, and the Nagelkerke R² was 0.363. CI, confidence interval; OR, odd ratio; B, Unstandardized Regression Coefficient; SE, Standard Error (of the coefficient); VD, vaginal delivery. Depth into the pelvis refers to the perpendicular distance from the pelvic inlet plane to the most inferior fetal/placental structure on midline sagittal MRI. Classification criteria: “Inferior” and “superior” positions of pelvic organs were defined using thresholds derived from the non-pregnant cohort (mean $\pm$ 1.96 SD). Bladder neck: “Inferior” threshold 14.87 mm. Cervix: “Inferior” threshold 11.21 mm. Posterior fornix: “Superior” threshold $>$47.58 mm.

Univariate analysis found that a deeper depth of the fetus and its appendages in the pelvis was positively associated with an “inferior” C position; the presence of a low-lying placenta and placenta previa was inversely related (Supplementary Table 3). However, multivariate logistic regression (Fig. 3B) showed that an “inferior” C position occurred in 71.8% of women with a depth of the fetus and its appendages in the pelvis $\ge$40 mm (OR = 11.46, 95% CI 2.53–51.92, p = 0.002) compared to 5.1% of those with a depth 20 mm. However, the low-lying placenta or placenta previa was not significantly associated with an “inferior” C position.

Univariate analysis showed that a higher BMI, advanced gestational age, deeper depth of the fetus and its appendages entering the pelvis, and the presence of a low-lying placenta or placenta previa were associated with a “superior” PF position (Supplementary Table 4). Multivariate logistic regression (Fig. 3C) revealed that with each additional week of gestational age, the odds of a “superior” PF position increase by 11% (p = 0.007). Moreover, a “superior” PF position occurred in 46.5% of women with entering pelvis depth 20 mm compared to 14.1% of those with depth $\ge$40 mm, representing 12.24 higher odds (95% CI 4.82–31.11, p 0.001). Furthermore, compared to women with normal placenta, those with the low-lying placenta and placenta previa had 2.96 (95% CI 1.18–7.39, p = 0.020) and 2.70 (95% CI 1.29–5.64, p = 0.008) higher odds of a “superior” PF, respectively.

This study presents a large MRI-based analysis of pelvic organ positional changes during pregnancy (n = 238), revealing distinct shifts relative to non-pregnant reference thresholds. Specifically, the BN and C showed inferior displacement, while the PF exhibited superior elevation. Fetal engagement depth was strongly associated with positional variations across all pelvic organs. A depth of $\ge$40 mm significantly increased the odds of inferior BN (OR = 5.04) and C positions (OR = 11.46), whereas a depth of 20 mm strongly predicted superior PF positioning (OR = 12.24).

Pregnancy is consistently associated with descent of pelvic organs [21]. In this study, the pregnant group showed inferior positioning of the BN and C, consistent with previous research [21], although with a modest reduction of 2.5–4.1 mm. Notably, the PF was superior in the pregnant group—an original finding not previously reported. Key factors associated with positional variation included the following.

First, gestational age (GA) was a significant factor. Compared with the non-pregnant group, the pregnant group had higher BMI values (25.7 vs. 21.8 kg/m²). However, PF was superior in the pregnant group. Of note, the mean BMI in the pregnant group was only slightly above the threshold for overweight. In the multicollinearity analysis, BMI was found to be correlated with GA, likely reflecting pregnancy-related weight gains due to uterine expansion and fetal growth. So, we included only GA in the final analysis. And we found that the BN position descended progressively with advancing pregnancy, whereas the C position did not show a significant association with GA. By contrast, PF elevation increased significantly with GA. Prior ultrasound studies suggest that the vaginal axis shifts upward and posteriorly as pregnancy progresses, which may support our findings [22]. Because pre-pregnancy BMI data were unavailable due to the retrospective nature of our study, future research incorporating pre-pregnancy BMI could offer more nuanced insights.

Fetal engagement depth emerged as a critical correlational factor, independent of GA. Women with deeper fetal engagement ($\ge$40 mm) had significantly higher odds of inferior BN and C positions (OR = 5.04 and OR = 11.46, respectively), while those with shallow engagement (20 mm) were 12.24 times more likely to exhibit superior PF positioning. This suggests that deeper engagement increases pressure on the pelvic floor, whereas shallow engagement may allow upward displacement of the PF. Our study is the first to integrate fetal engagement depth into pelvic floor modeling, distinguishing it from GA, that serves as a temporal rather than mechanical indicator.

We also examined fetal presentation and placental location. While breech presentation and placenta previa were initially associated with inferior BN and C positions, these associations lost significance after adjusting for confounders. However, a low-lying placenta and placenta previa remained significant predictors of superior PF elevation (OR = 2.96 and OR = 2.70, respectively). Given that 51.3% of our sample had these conditions, they likely contributed to the observed superior PF elevation.

Vaginal delivery history was positively associated with BN descent, consistent with prior studies on post-childbirth pelvic floor dysfunction [1, 3, 16, 23]. However, no significant association was found between vaginal delivery history and C or PF positioning, possibly because of confounding factors or characteristics of the study population.

Finally, maternal posture during MRI is another critical, though often overlooked, factor influencing pelvic floor measurements. While most MRIs in our hospitals are performed in the supine position, lateral decubitus positioning may be used in later pregnancy to avoid fetal distress. Because this was a retrospective study, we lacked precise data on patient positioning during scans. Previous pelvic floor imaging studies have used varied postures. For example, Chan et al. [21] used ultrasound in the supine position, and Oliphant et al. [22] used ultrasound in the lithotomy position. Routzong et al. [15, 16] conducted MRI studies in the lateral decubitus position for gravid patients, while Yagi et al. [17] performed MRIs in the supine position at rest. These differences in imaging posture can affect the pelvic load distribution [24], highlighting an important area for further investigation.

Overall, our study underscores the importance of a comprehensive approach when assessing pelvic anatomy during pregnancy. Factors such as GA, fetal engagement depth, fetal presentation, and placental location must all be considered. Because pelvic organ positioning likely differs between upright and lying postures—and both are physiologically relevant—future research should also investigate pelvic anatomy in standing positions. Our study offers new perspectives and valuable insights into evaluating pelvic floor anatomy during pregnancy.

Several limitations should be acknowledged. First, because of potential teratogenic risks, MRI was not performed during the first trimester. Second, because this was a retrospective study, causality cannot be definitively established, and the identified associations should be interpreted with caution. Third, selection bias may arise from the restriction of our sample to pregnant women undergoing clinically indicated MRI, and stringent inclusion/exclusion criteria. Furthermore, the single-center design with limited ethnic diversity may constrain generalizability to broader pregnant populations. Fourth, the wide confidence intervals of the odds ratios for some factors reflect limited sample sizes in the subgroup analysis. Fifth, the pelvic floor is a three-dimensional structure, and two-dimensional measurements or single reference lines may not fully capture its anatomical changes during pregnancy. Future research should include prospective, large-scale studies across diverse racial and cultural groups to improve generalizability. More representative markers for the posterior compartment, such as the anorectal angle, the use of pelvic inclination correction system [25] and three-dimensional MRI are also recommended. Additionally, incorporating intra- and inter-observer variability assessments would enhance measurement reliability—an acknowledged limitation of this study.

Despite these limitations, the study has several strengths. We systematically explored pelvic organ positional variations in relation to obstetric factors—a novel approach in this field. While an ideal study design would involve serial MRIs before and throughout pregnancy, such an approach raises ethical concerns due to repeated MRI exposure. This study provides a large-scale dataset that offers important insights into pelvic floor changes during pregnancy. To define “inferior” or “superior” organ positions, we established reference thresholds based on a large non-pregnant control group, enhancing the robustness and clinical relevance of our findings.