Nomogram model for predicting invasive placenta in patients with placenta previa: integrating MRI findings and clinical characteristics

This study aims to validate a nomogram model that predicts invasive placenta in patients with placenta previa, utilizing MRI findings and clinical characteristics. A retrospective analysis was conducted on a training cohort of 269 patients from the Second Affiliated Hospital of Fujian Medical University and a validation cohort of 41 patients from Quanzhou Children’s Hospital. Patients were classified into noninvasive and invasive placenta groups based on pathological reports and intraoperative findings. Three clinical characteristics and eight MRI signs were collected and analyzed to identify risk factors and develop the nomogram model. The mode’s performance was evaluated in terms of its discrimination, calibration, and clinical utility. Independent risk factors incorporated into the nomogram included the number of previous cesarean sections ≥ 2 (odds ratio [OR] 3.32; 95% confidence interval [CI] 1.28–8.59), type-II placental bulge (OR 17.54; 95% CI 3.53–87.17), placenta covering the scar (OR 2.92; CI 1.23–6.96), and placental protrusion sign (OR 4.01; CI 1.06–15.18). The area under the curve (AUC) was 0.908 for the training cohort and 0.803 for external validation. The study successfully developed a highly accurate nomogram model for predicting invasive placenta in placenta previa cases, based on MRI signs and clinical characteristics.


Research subjects
This study received approval from the institutional review boards of both the Second Affiliated Hospital of Fujian Medical University and Quanzhou Children's Hospital (Reference No. 463).Due to its retrospective nature, written informed consent was waived, as approved by the Ethics Committees of the two hospitals.The study adhered to the guidelines of the Declaration of Helsinki and was conducted in compliance with relevant guidelines and regulations.
The sample size for our clinical prediction model was determined using Riley's methods, employing the "pmsampsize" package in R 12 .Based on our prior clinical observations, the incidence rate of PAS ranges between 20 and 25%.Consequently, we adopted a midpoint rate of 22% for our calculations.This approach indicated a minimum required sample size of 264, assuming the inclusion of five predictive factors.Additionally, in developing the model, we adhered to the guideline of ensuring a minimum of 10 events per predictive factor.
Placenta previa is increasingly recognized as a risk factor for PAS.Patients suspected of placenta previa, identified via ultrasound, are typically recommended for further MRI scans to exclude PAS 13,14 .From January 2014 to December 2020, we included 798 patients from the Second Affiliated Hospital of Fujian Medical University who underwent MRI scans for placenta previa as identified by ultrasound.The inclusion criteria were: (1) placenta previa confirmed by MRI after 28 weeks of gestation; (2) prenatal ultrasound and MRI conducted in our hospital; (3) singleton pregnancies; (4) cesarean section deliveries and any related surgeries performed in our hospital; (5)  absence of prior hysteroscopic surgery; (6) the last MRI assessment of the placenta conducted within two weeks before cesarean section.Exclusion criteria were: (1) poor quality images or severe motion artifacts; (2) patients with in vitro fertilization due to the potential higher risk of invasive placenta; (3) incomplete data.Ultimately, 269 patients were included as a training cohort, with relevant clinical and radiological data retrospectively collected.Similarly, 158 patients from Quanzhou Children's Hospital during the same period were considered, with 41 patients eventually enrolled as a validation cohort.Clinical and radiological data were collected and compared between the cohorts.The process is illustrated in Fig. 1.

Instruments and equipment
MRI scans were performed using a 1.5 T MR scanner (Achieva, Philips Medical System, Netherlands) equipped with a body array coil covering the entire pelvis.To facilitate optimal evaluation of the bladder-serosal interface, patients were instructed to maintain a full bladder during the scan.The scanning sequences included ssh_TSE,

MRI image evaluation
The MRI images were reviewed by two experienced radiologists, with 10 and 15 years of experience in pelvic MRI diagnostics, respectively.Prior to the study, these radiologists underwent training to standardize the interpretation of MRI images.They were blinded to patients' clinical histories, previous radiological reports, and surgical and pathological outcomes.In cases of disagreement, a consensus was reached by accepting the positive diagnosis made by either radiologist.

PAS grading and patient classification
According to the FIGO, the PAS grading diagnosis is confirmed by pathological reports and intrapartum findings 17 .Patients were categorized into noninvasive placenta groups (including PAS grade 1 and non-adherent placenta) and invasive placenta groups (including PAS grades 2 and 3).The diagnostic criteria and classification of noninvasive and invasive placenta are outlined in Table 2.All patients in this study underwent cesarean section delivery, with PAS grading determined through a combination of intraoperative findings and pathological results.All suspected cases of placental implantation were routinely referred for pathological analysis.The interpretation of intraoperative findings and pathological diagnoses was overseen by an obstetrician with over 20 years of experience, while the pathological examination was conducted by a pathologist with more than 15 years of experience.

Statistical analysis
Statistical analyses were performed using SPSS version 23.0 (SPSS, Inc, Chicago, IL, USA) and R software (R version 4.0.3,The R Project for Statistical Computing, www.r-proje ct.org).Continuous variables were presented as mean ± standard deviation, while categorical variables were expressed as numbers (percentages).The Chi-square test was employed for analyzing categorical variables.Multivariate logistic regression analysis was conducted using significant risk factors identified.The variance inflation factor (VIF) was calculated to assess collinearity, with a VIF < 5 indicating no significant collinearity.A p value < 0.05 was considered statistically significant.
A nomogram was created using the coefficients from the logistic regression model, weighted in the R package "rms".The discriminative performance of the model was assessed by the area under the receiver operating characteristic (ROC) curve.Calibration was evaluated using unreliability statistics to generate a calibration curve.Bootstrapping with 1000 samples was performed to assess internal calibration and discrimination.The clinical utility of the nomogram was evaluated using decision curve analysis (DCA), determining the net benefit at various threshold probabilities.

Clinical and radiological variables analysis
The clinical and radiological characteristics of the patients are summarized in Table 3.In the training cohort of 269 patients, 180 were classified into the noninvasive group and 89 into the invasive group.The distribution was as follows: 152 patients (56.5%) were nonadherent, 28 (10.4%) were PAS grade 1, 60 (22.3%) were PAS grade 2, and 29 (10.8%) were PAS grade 3.No significant differences were observed in age and number of abortions between the groups.However, significant differences were found in the number of previous cesarean sections ≥ 2 and various MRI signs, including intraplacental T2 dark bands, placental bulge type I, placental bulge type II, uterine serosal hypervascularity, abnormal intraplacental vascularity and/or bridging vessels, placental recess, placenta covering scar, and placental protrusion sign (Table 3).Multivariable logistic regression analysis identified the number of previous cesarean sections ≥ 2, placental bulge type-II, placenta covering scar, and placental protrusion sign as independent risk factors for invasive placenta (Table 4).The analysis indicated increased risks for invasive placenta by factors of 3.32, 17.57, 2.94, and 4.01, respectively, for these variables.No significant collinearity was found among the significant predictors (Table 5).

Intraoperative findings
Pathological results

Non-adherent
The placenta can be peeled off on its own Grade 1 The placenta cannot be peeled off on its own, it needs to be peeled off by hand.The placental detach surface is rough and bleeding after stripping The placental villi are attached to the myometrium Grade 2 Abnormal dilated placental blood vessels are seen during the operation.The placenta cannot be detached by itself and the placenta is torn by hand, which required oval forceps to remove or cut off the implanted part of the myometrium The villi can be seen invading the myometrium

Validation cohort
The validation cohort from Quanzhou Children's Hospital comprised 41 patients.There were no significant differences in relevant variables between the two cohorts (Table 6).

Nomogram model creation and evaluation
Utilizing the coefficients from the final model, a nomogram was developed (Fig. 4).The first line on the nomogram provides reference scores for each risk factor.Lines 2-5 represent the scores assigned to each individual risk factor, including the number of previous cesarean sections ≥ 2, placental bulge type-II, placenta covering scar, and placental protrusion sign.The sixth line denotes the total score, calculated by summing the scores of each risk factor.The final line on the nomogram represents the predicted probability of invasiveness as determined by the model.The model's discriminative ability was assessed using ROC curves.The AUC was 0.908 for the training cohort (Fig. 5a) and 0.803 for the validation cohort (Fig. 5b), indicating good predictive performance.The optimal cutoff values were 0.703 for the training cohort and 0.545 for the validation cohort, with corresponding sensitivities of 80.9% and 54.5%, and specificities of 89.4% and 100%.
Calibration of the model was performed, as depicted in Fig. 6a, b.The prediction curve closely matched the standard curve, with an average error less than 0.05, demonstrating accurate calibration.DCA, shown in  Fig. 7, was conducted to evaluate the model's clinical utility.The analysis indicated that using the nomogram for predicting invasive placenta probability is more beneficial than treating all or none of the patients when the threshold probability is greater than 8%.

Discussion
Nomograms serve as simplified predictive models, presenting visual results in a graphical format.They are extensively employed in the prediction of tumorigenesis and prognosis for various cancers, including breast, gastric, and prostate cancers [18][19][20] .Prior research has primarily focused on the association between MRI signs and the occurrence of PAS [21][22][23][24][25][26][27] .In this study, we developed and validated a nomogram for predicting PAS in patients with placenta previa, incorporating clinical features and MRI signs.The nomogram, comprising the number of previous cesarean sections ≥ 2, placental bulge type-II, placenta covering scar, and placental protrusion sign, proved to be a practical and effective tool for clinical decision-making.
In our study, age and the number of previous abortions did not significantly differ between the noninvasive and invasive groups, aligning with Baughman's findings 28 .The FIGO guideline suggests that, in comparison to cesarean sections, induced abortions have a minimal impact on the myometrium of the uterus, thereby posing a lower risk of invasive placenta.Our research identified that having two or more prior cesarean sections was an  www.nature.com/scientificreports/independent risk factor for PAS.This is attributed to the disruption of endometrial decidualization over uterine scarring areas, allowing villous trophoblasts to abnormally anchor deeply into the myometrium.This leads to uncontrolled invasion by extra-villous trophoblasts, a characteristic feature of PAS 29 .While numerous MRI features of PAS have been extensively described in literature [21][22][23][24][25][26][27] , and many studies have demonstrated a strong correlation between MRI findings and intraoperative observations of invasive placentas 6,16,21 , certain MRI signs were excluded from our study.These include myometrial thinning, interrupted myometrium, heterogeneous placenta, abnormal uterine bulging, loss of the placental-myometrial interface, and percretism signs.The signs of interrupted myometrium and myometrial thinning, characterized by abrupt interruption or thinning of the myometrium to less than 1 mm at the site of focal placental bulging, often cooccur with placental bulging.Myometrial thinning, in particular, can lead to significant observational bias 30 .The heterogeneous placenta sign, indicating varied signal intensity within the placenta, is highly influenced by placental maturity and various other factors.Abnormal uterine bulging, manifested as widening of the lower uterine segment resembling an hourglass shape, lacks a precise definition in existing studies and is often confounded with focal placental bulge.The loss of the placental-myometrial interface becomes challenging to discern in cases of thin myometrium during late pregnancy.Percretism, indicating direct invasion into adjacent organs, has a notably low sensitivity.In our experience, the bridging vessels sign, often observed in type I PAS, can be difficult to distinguish from the sign of placental hypervascularity.As a result, we combined the bridging vessels sign with the placental hypervascularity sign for analysis in this study.
In this study, we focused on eight prominent and practical MRI signs, discovering that placental bulge type II, placenta covering scar, and placental protrusion were independent risk factors for invasive placenta.Jha et al. reported that the presence of a placental bulge, when combined with other indicators of invasive placenta, achieved a predictive ability of 100% 31 .Leyendecker similarly identified outward placental bulging as the most indicative sign 22 .Consistent with these findings, our study confirmed its independent risk factor status, with a sensitivity of 57.3% and a specificity of 98.3%.The sign of placenta covering scar, often associated with previous cesarean sections, demonstrated high sensitivity (83.10%) and specificity (75.00%) in our study.The placental protrusion sign, indicative of uncontrolled villous proliferation in the absence of a regular basal plate, showed a high specificity of 97.8%, although the sensitivity was only 44.90%.This aligns with Ueno's findings, which associated this sign with invasive placenta praevia 32 .Our results reinforce the importance of these MRI signs as valuable predictors in the assessment of invasive placenta risk.
The model's efficacy was validated in an external cohort, focusing on its discrimination, calibration, and clinical applicability.This validation offers a foundation for developing personalized treatment strategies for patients prior to surgery.The model demonstrated robust predictive accuracy, with an AUC of 0.908 in the training cohort and 0.803 in the validation cohort.These results affirm the model's effectiveness, substantiated through external validation.
In 2021, Li developed and validated the first nomogram for PAS, based on a retrospective single-center study involving 118 patients in the training cohort and 65 in the validation cohort, drawn from different periods 11 .Our study encompasses a larger training cohort.A key distinction between our study and Li's lies in the selection of MRI signs.Li's study included signs like focal myometrial thinning less than 1 mm, focal myometrial interruption, intraplacental T2 dark bands, hemorrhage within the placenta, disorganized abnormal placental vascularity, abnormal uterine bulging, and loss of the double-line sign.We noted previously that the precise definition of uterine bulging is unclear and often confounded with placental bulging.Placental bulging is a frequent observation, typically coinciding with signs of myometrial thinning, interruption, and loss of the placental-myometrial interface.Therefore, we opted for Chen's more intuitive and observer-stable classification of placental bulge 16 .This sign demonstrated the highest value in our nomogram.Additionally, the sensitivity of the newly introduced loss of double-line sign by Li was only 55.9%.Our nomogram achieved an AUC of 90.8%, surpassing the performance of Li's nomogram.

Figure 2 .
Figure 2. 39 yrs old, PAS grade 3. A and B: the red arrow show the sign of placental bulge type-II on the OCOR (a) and OSAG (b) BTFE sequence; (c): the yellow arrow show the sign of uterine serosal hypervascularity, the green arrow show sign of bridging vessels on OSAG ssh_TSE sequence; (d): the red arrow show the sign of placental protrusion on OSAG BTFE sequence.

Figure 3 .
Figure 3.A and B are 34 yrs old, PAS grade 3. A:the red arrow show the sign of intraplacental T2 dark bands on the OSAG BTFE sequence; B: the red arrow show the sign of placental recess on the OCOR ssh_TSE sequence; (c) and (d) are 29 yrs old, PAS grade 2. The red arrow show the sign of placental bulge type-I on the OSAG (a) and OCOR (b) BTFE sequence.

Figure 4 .
Figure 4. Nomogram model for predicting invasive placenta.For a given patient, use the axis at the top of the figure and points are assigned to each variable and get the total score.The total score corresponds to a predicted probability of invasive placenta.

Figure 5 .
Figure 5. Receiver operating characteristic curve (ROC) for evaluating the discriminative performance of training cohort (a) and validation cohort (b).

Figure 6 .
Figure 6.The calibration curve of the nomogram the in the training set (a) and validation set (b).

Figure 7 .
Figure 7. Decision curve analysis of nomogram model in a validation cohort.
15,16radiologists contributed equally to the data used in constructing the nomogram model.Key MRI features indicative of abnormal placentation, as identified in the literature, were assessed and Placental protrusion sign15,16.These signs are detailed in Table1, with examples illustrated in Figs.2 and 3. Additionally, patient age, number of abortions and cesarean sections, gestational age at MRI, and gestational age at delivery were recorded.

Table 1 .
The definition of MR signs.

Table 2 .
Diagnostic criteria and classification of noninvasive placenta and invasive placenta.

Table 3 .
The clinical and radiological variables of patients of training cohort.

Table 4 .
Significant predictors of the invasive placenta from multivariate logistic regression analysis.

Table 5 .
Collinearity of significant predictors of invasive placenta.

Table 6 .
Comparison of significant predictors between the training and validation cohorts.