The diagnostic value of multiple schemes of MRI standard features combined with risk factors and ultrasound for Placenta accreta spectrum

A total of 199 participants were included in the study, with an average age of 31.75 ± 4.27 years (range: 19–44 years). The PAS group comprised 89 participants (placenta accreta 34(17.1), placenta increta 17(8.5), placenta percreta 38(19.1)) with an average age of 32.69 ± 4.12 years, while the non-PAS group had 110 participants with an average age of 30.99 ± 4.26 years. The difference in average age between the two groups is less than 2 years. There were no statistically significant differences in other general data (see Table  1 ). Table 1 Comparison of PAS risk factors between groups based on FIGO guidelines Variable Non PAS group ( n  = 110) PAS group ( n  = 89) Statistic P Degree of education Junior high school and below 14(12.7%) 23(25.8%) 6.148 0.188 High school and vocational school 19(17.3%) 14(15.7%) junior college 33(30.0%) 19(21.3%) undergraduate college 37(33.6%) 28(31.5%) Master’s degree or above 7(6.4%) 5(5.6%) Menstrual cycle normal 98(89.1%) 84(94.4%) 1.763 0.184 abnormal 12(10.9%) 5(5.6%) History of dysmenorrhea No 100(90.9%) 86(96.6%) 2.636 0.104 Yes 10(9.1%) 3(3.4%) History of gestational trophoblastic disease/neoplasia No 108(98.2%) 87(97.8%)  0.999 Yes 2(1.8%) 2(2.2%) History of Scar Pregnancy No 109(99.1%) 88(98.9%) > 0.999 Yes 1(0.9%) 1(1.1%) History of missed abortion No 104(94.5%) 81(91.0%) 0.940 0.332 Yes 6(5.5%) 8(9.0%) History of biochemical pregnancy No 109(99.1%) 87(97.8%) 0.034 0.853 Yes 1(0.9%) 2(2.2%) History of ectopic pregnancy No 108(98.2%) 88(98.9%)  0.999 Yes 2(1.8%) 1(1.1%) History of spontaneous abortion No 102(92.7%) 85(95.5%) 0.670 0.413 Yes 8(7.3%) 4(4.5%) History of vaginal delivery No 88(80.0%) 75(84.3%) 0.605 0.437 Yes 22(20.0%) 14(15.7%) History of surgical termination of pregnancy No 65(59.1%) 48(53.9%) 0.533 0.465 Yes 45(40.9%) 41(46.1%) History of cervical dilation and curettage No 69(62.7%) 49(55.1%) 1.199 0.273 Yes 41(37.3%) 40(44.9%) History of myomectomy No 107(97.3%) 89(100.0%) 0.97 0.325 Yes 3(2.7%) 0(0.0%) History of endometrial resection No 110(100.0%) 88(98.9%) 0.447 Yes 0(0.0%) 1(1.1%) In vitro fertilization No 104(94.5%) 88(98.9%) 1.593 0.207 Yes 6(5.5%) 1(1.1%) Proportion was calculated by column Comparison of PAS risk factors between groups based on FIGO guidelines Proportion was calculated by column In this study, all PAS patients were diagnosed according to the FIGO guidelines, combining intraoperative findings and pathological diagnosis. All delivered placentas underwent pathological examination. Among them, histopathological examination of the uterus was performed on seven patients who underwent hysterectomy and on one patient who underwent partial uterine resection. The histological and intraoperative findings were consistent with the pathological diagnosis of placental disorders. Risk factors were compared between groups according to the FIGO guidelines. The results showed that the proportion of three risk factors, including a history of CD, pregnancy with placenta previa, and advanced maternal age pregnancy, was significantly higher in the PAS group (see Table  1 ). In the PAS group, there were 67 cases (75.28%) 67 cases (75.28%) and 32 cases (35.96%) with a history of CD, placenta previa, and advanced maternal age pregnancy, respectively. In contrast, there were only 44 cases (40.00%) 31 cases (28.18%) and 17 cases (15.45%) for these conditions in the non-PAS group, respectively. A comparison of the 11 commonly used standardized MRI features revealed that, except for the feature of Focal exophytic mass, there were statistically significant differences in the other 10 features between the PAS group and the non-PAS group (refer to Table 2 ; Fig. 3 ). Table 2 Comparison of MRI standard features for diagnosing PAS between groups Variable Non PAS group ( n  = 110) PAS group ( n  = 89) χ 2 P T2-dark intraplacental bands No 110(100.0%) 35(39.3%) 91.597 < 0.001 Yes 0(0.0%) 54(60.7%) Placental/uterine bulge No 110(100.0%) 57(64.0%) 47.129 < 0.001 Yes 0(0.0%) 32(36.0%) Loss of retroplacental T2-hypointense line No 109(99.1%) 15(16.9%) 141.673 < 0.001 Yes 1(0.9%) 74(83.1%) Myometrial thinning No 108(98.2%) 15(16.9%) 137.847 < 0.001 Yes 2(1.8%) 74(83.1%) Bladder wall interruption No 110(100.0%) 82(92.1%) 6.799 0.009 Yes 0(0.0%) 7(7.9%) Focal exophytic mass No 110(100.0%) 87(97.8%) 0.199 Yes 0(0.0%) 2(2.2%) Abnormal vascularization of placental bed No 17(15.5%) 5(5.6%) 4.841 0.028 Yes 93(84.5%) 84(94.4%) Placental heterogeneity No 103(93.6%) 45(50.6%) 47.89 < 0.001 Yes 7(6.4%) 44(49.4%) Asymmetric shape/thickening of the placenta No 110(100.0%) 70(78.7%) 25.962 < 0.001 Yes 0(0.0%) 19(21.3%) Placental ischemic infarction No 110(100.0%) 72(80.9%) 22.974 < 0.001 Yes 0(0.0%) 17(19.1%) Abnormal intraplacental vascularity No 108(98.2%) 72(80.9%) 17.016 < 0.001 Yes 2(1.8%) 17(19.1%) Proportion was calculated by column Comparison of MRI standard features for diagnosing PAS between groups Proportion was calculated by column Fig. 3 The standard MRI features from pregnant women. Axial ( A ), coronal ( B ), and sagittal ( C ) MRI images depict normal placentas from a 34-year-old pregnant woman at 32 weeks of gestation. D A 35-year-old woman at 31 weeks of gestation, with coronal view showing Abnormal intraplacental vascularity (white arrow). E A 32-year-old woman at 32 weeks of gestation. The coronal T2WI demonstrates multiple prominent, tortuous, and enlarged vascular flow voids with low signal intensity within the placental bed (white arrow), accompanied by loss of the placenta-myometrium interface. The abnormal vasculature extends into the myometrium and serosa, resulting in indistinct delineation of the myometrial and serosal layers. F A 23-year-old woman at 28 weeks of gestation, with axial view demonstrating Placental heterogeneity, including patchy high signal and stripes of low signal (white arrow) within the placenta. G A 31-year-old woman, at 36 weeks of gestation, with axial view showing a retroplacental T2-hypointense line on the posterior uterine wall (yellow arrow), and T2-dark intraplacental bands on the anterior uterine wall (white arrow). H 36y, at 33 weeks of gestation; the sagittal T2WI sequence demonstrates Placental/uterine bulge (white arrow). I 31y, at 36 weeks of gestation; the sagittal T2WI sequence demonstrates Placental ischemic infarction (white arrow). The normal placenta demonstrates uniform thickness with smooth, gradually tapering margins and no areas of deep tissue defect. Sites of placental chronic infarction exhibit asymmetric thinning and appear hypotintense on T2-weighted imaging. J 31y, at 33 weeks of gestation; the axial T2WI sequence shows Myometrial thinning (white arrow). K 32y, at 36 weeks of gestation, the sagittal view reveals Bladder wall interruption (white arrow). L 35y, at 39 weeks of gestation, the sagittal view demonstrates Asymmetric shape/thickening of the placenta (white arrow). M From the same patient as 2 J, the axial view shows irregular T2-dark intraplacental bands (white arrow). N - O 37y, at 32 weeks of gestation, the coronal and axial view reveals Focal exophytic mass, the mass protrudes into the bladder (white arrow) The standard MRI features from pregnant women. Axial ( A ), coronal ( B ), and sagittal ( C ) MRI images depict normal placentas from a 34-year-old pregnant woman at 32 weeks of gestation. D A 35-year-old woman at 31 weeks of gestation, with coronal view showing Abnormal intraplacental vascularity (white arrow). E A 32-year-old woman at 32 weeks of gestation. The coronal T2WI demonstrates multiple prominent, tortuous, and enlarged vascular flow voids with low signal intensity within the placental bed (white arrow), accompanied by loss of the placenta-myometrium interface. The abnormal vasculature extends into the myometrium and serosa, resulting in indistinct delineation of the myometrial and serosal layers. F A 23-year-old woman at 28 weeks of gestation, with axial view demonstrating Placental heterogeneity, including patchy high signal and stripes of low signal (white arrow) within the placenta. G A 31-year-old woman, at 36 weeks of gestation, with axial view showing a retroplacental T2-hypointense line on the posterior uterine wall (yellow arrow), and T2-dark intraplacental bands on the anterior uterine wall (white arrow). H 36y, at 33 weeks of gestation; the sagittal T2WI sequence demonstrates Placental/uterine bulge (white arrow). I 31y, at 36 weeks of gestation; the sagittal T2WI sequence demonstrates Placental ischemic infarction (white arrow). The normal placenta demonstrates uniform thickness with smooth, gradually tapering margins and no areas of deep tissue defect. Sites of placental chronic infarction exhibit asymmetric thinning and appear hypotintense on T2-weighted imaging. J 31y, at 33 weeks of gestation; the axial T2WI sequence shows Myometrial thinning (white arrow). K 32y, at 36 weeks of gestation, the sagittal view reveals Bladder wall interruption (white arrow). L 35y, at 39 weeks of gestation, the sagittal view demonstrates Asymmetric shape/thickening of the placenta (white arrow). M From the same patient as 2 J, the axial view shows irregular T2-dark intraplacental bands (white arrow). N - O 37y, at 32 weeks of gestation, the coronal and axial view reveals Focal exophytic mass, the mass protrudes into the bladder (white arrow) Multivariate analysis was conducted on 10 standardized MRI features that showed differences between the non-PAS and PAS groups with logistic regression. Subsequently, MRI indicators with statistical significance were identified. Due to the low frequency of certain feature categories and the resulting large standard errors of estimation, the bootstrap method was employed to estimate effect sizes following stepwise selection. The results indicated that three MRI features: T2-dark intraplacental bands, Loss of retroplacental T2-hypointense line, and Asymmetric shape/thickening of the placenta were statistically significant indicators (Table  3 ). These three features were retained as the key MRI indicators in the final diagnostic model. Table 3 Results of multivariate logistic regression bootstrap analysis efficiency MRI features B Bias SE P T2-dark intraplacental bands 19.944 -1.070 4.343 0.001 Loss of retroplacental T2-hypointense line 5.151 6.156 8.345 0.001 Asymmetric shape/thickening of the placenta 20.137 -3.809 7.791 0.022 Results of multivariate logistic regression bootstrap analysis efficiency The results indicated that the AUC for diagnosing PAS using risk factors, US, and MRI individually was greater than 0.8 (Table  4 ). Each demonstrated diagnostic efficacy. However, MRI demonstrated the highest diagnostic efficacy, while risk factors had the lowest. Combining MRI with other methods significantly enhanced diagnostic efficiency (Table  4 ; Fig.  4 ). There was no significant difference in diagnostic efficacy among the MRI+Risk Factors, MRI + US, and MRI + US+Risk Factors strategies. Table 4 Comparison of efficacy of multiple schemes in diagnosing PAS Diagnostic indicators AUC(95%CI) SE P Sensitivity Specificity Youden index Risk Factors a 0.835(0.778, 0.891) 0.029 < 0.001 0.618 0.906 0.524 US b 0.895(0.850, 0.940) 0.023 < 0.001 0.798 0.849 0.647 MRI b 0.931(0.895, 0.967) 0.022 < 0.001 0.865 0.991 0.856 US+Risk Factors b 0.899(0.855, 0.943) 0.023 < 0.001 0.843 0.811 0.654 MRI+Risk Factors c 0.971(0.947, 0.996) 0.013 < 0.001 0.933 0.934 0.867 MRI + US c 0.975(0.956, 0.993) 0.010 < 0.001 0.876 0.981 0.857 MRI + US+Risk Factors c 0.974(0.953, 0.995) 0.011 < 0.001 0.910 0.972 0.882 There was no statistically significant difference in the AUC values between US and MRI. However, the specificity of MRI was significantly higher than that of US, with a statistically significant difference ( P  < 0.05) The same letters indicate no significant difference between groups, while different letters indicate significant differences. The statistically significant factors of CD history, pregnancy with placenta previa, and advanced maternal age pregnancy (Table  1 ) were used as risk factor indicators. When US is included in the diagnostic combination, only advanced maternal age pregnancy is considered as a risk factor, since CD history and pregnancy with placenta previa are already incorporated into the US scoring calculation Comparison of efficacy of multiple schemes in diagnosing PAS There was no statistically significant difference in the AUC values between US and MRI. However, the specificity of MRI was significantly higher than that of US, with a statistically significant difference ( P  < 0.05) The same letters indicate no significant difference between groups, while different letters indicate significant differences. The statistically significant factors of CD history, pregnancy with placenta previa, and advanced maternal age pregnancy (Table  1 ) were used as risk factor indicators. When US is included in the diagnostic combination, only advanced maternal age pregnancy is considered as a risk factor, since CD history and pregnancy with placenta previa are already incorporated into the US scoring calculation Fig. 4 Comparison of area under the ROC curve of various schemes for diagnosis of PAS (Risk Factors: RFs) Comparison of area under the ROC curve of various schemes for diagnosis of PAS (Risk Factors: RFs) For the three diagnostic schemes (MRI+Risk Factors, MRI + US, and MRI + US+Risk Factors) with no differences detected by AUC, the scheme/model improvement evaluation method was further used for comparison, which was achieved by comparing NRI and IDI (Table  5 ). The results showed that the diagnostic efficiency of prenatal PAS ranked from high to low as: MRI + US+Risk Factors, MRI +Risk Factors, MRI + US. Table 5 NRI and IDI compared with different diagnostic schemes Reference scheme Comparison scheme NRI(Continuous) IDI MRI+Risk Factors MRI + US -0.583(-0.839, -0.328) P  < 0.001 -0.023(-0.054, 0.007) P  = 0.129 MRI+Risk Factors MRI + US+Risk Factors -0.2103(-0.477, 0.056) P  = 0.122 -0.013(-0.041, 0.014) P  = 0.352 MRI + US MRI + US+Risk Factors 0.537(0.309, 0.766) P <  0.001 0.011 (-0.002, 0.023) P  = 0.089 All IDI values have no statistical significance. Although AUC differences were not statistically significant, NRI suggested improved reclassification when risk factors were added to MRI + US, supporting incremental value in risk stratification NRI and IDI compared with different diagnostic schemes -0.583(-0.839, -0.328) P  < 0.001 -0.023(-0.054, 0.007) P  = 0.129 -0.2103(-0.477, 0.056) P  = 0.122 -0.013(-0.041, 0.014) P  = 0.352 0.537(0.309, 0.766) P <  0.001 0.011 (-0.002, 0.023) P  = 0.089 All IDI values have no statistical significance. Although AUC differences were not statistically significant, NRI suggested improved reclassification when risk factors were added to MRI + US, supporting incremental value in risk stratification

This retrospective study included 203 pregnant women from November 2018 to December 2022. All participants delivered at Zhengzhou Central Hospital with complete US and MRI. Three cases of multiple pregnancy and one stillbirth were excluded, ultimately 199 patients were included in the analysis. Inclusion criteria were: (1) clinical patients suspected of PAS or with a posteriorly located placenta; (2) the interval between MRI and ultrasound examinations is less than 2 weeks; (3) complete clinical records from the perinatal period were available. Exclusion criteria were: (1) twin or multiple pregnancies; (2) incomplete clinical or imaging data; (3) poor image quality or incomplete MRI sequences; (4) missing delivery outcomes or unavailable surgical/pathological confirmation due to loss to follow-up; (5) stillbirth. Ultimately 199 patients were included in the analysis (The flowchart of study subject selection is shown in Fig.  1 ). This study was approved by the ethics committee of the Zhengzhou Central Hospital Affiliated to Zhengzhou University (approval number: 202241). Given the retrospective nature of the study and anonymization of patient data, informed consent was waived. Fig. 1 The flowchart of study subject selection The flowchart of study subject selection PAS diagnosis was based on intraoperative findings as the primary basis, combined with histological examination as a supplementary confirmation, and obstetricians made diagnoses according to the International Federation of Gynecology and Obstetrics (FIGO) standards [ 9 , 10 ]. According to the existence of PAS, the subjects were divided into PAS group and non-PAS group. The risk factors of PAS were collected and recorded according to FIGO’s risk factors classification and types by consulting the medical records [ 11 ]. Transabdominal and transvaginal ultrasound examinations were performed using a GE Voluson E8 color Doppler ultrasound system. The US diagnosis was made according to the PAS US sign rating scale [ 12 , 13 ] (Supplementary Tables 1 and Fig. 2 ). The final diagnosis was made after a preliminary scoring by a US diagnostician and review by a senior radiologist. Fig. 2 Ultrasonographic images of normal placenta and placenta accreta spectrum disorders. A A 37-year-old woman at 39 weeks of gestation, showing normal placental location and thickness. The white arrow indicates a continuous hypoechoic retroplacental space, with no increased vascularity at the placental base. B A 35-year-old woman at 33 weeks of gestation. The white arrow indicates a continuous hyperechoic bladder line. C A 37-year-old woman at 39 weeks of gestation. The white arrow demonstrates regular vascular signals at the placental base. D A 37-year-old woman at 39 weeks of gestation, showing an intact cervical morphology with normal vascular distribution. E A 32-year-old woman at 32 weeks of gestation. The white arrow indicates loss of hypoechoic retroplacental space. The yellow arrow points to hypoechoic or anechoic placental lacunae. F A 32-year-old woman at 32 weeks of gestation with placenta previa. The white arrow indicates loss of hypoechoic retroplacental space. G A 28-year-old woman at 27 weeks of gestation. The white arrow indicates interruption of the continuous hyperechoic bladder line. H A 35-year-old woman at 29 weeks of gestation with placenta previa. The white arrow shows confluent areas of hypoechoic or anechoic placental lacunae. I A 31-year-old woman at 33 weeks of gestation. The yellow arrow demonstrates increased, clustered vascular signals at the placental base. J A 35-year-old woman at 29 weeks of gestation with placenta previa. The white arrow indicates increased vascularity at the placental base. The yellow arrow shows the presence of anechoic placental lacunae. The red arrow indicates blood sinuses of the cervix. K A 28-year-old woman at 27 weeks of gestation. The white arrow indicates bridging vessels at the placental base. The red arrow shows coalescent patches of cervical blood sinuses. L A 29-year-old woman at 27 weeks of gestation with placenta previa. The white arrow indicates bridging vessels and the yellow arrow shows blood sinuses of the cervix. (PL: Placenta; UA: Umbilical Artery; CX: Uterine Cervix; BL: Bladder) Ultrasonographic images of normal placenta and placenta accreta spectrum disorders. A A 37-year-old woman at 39 weeks of gestation, showing normal placental location and thickness. The white arrow indicates a continuous hypoechoic retroplacental space, with no increased vascularity at the placental base. B A 35-year-old woman at 33 weeks of gestation. The white arrow indicates a continuous hyperechoic bladder line. C A 37-year-old woman at 39 weeks of gestation. The white arrow demonstrates regular vascular signals at the placental base. D A 37-year-old woman at 39 weeks of gestation, showing an intact cervical morphology with normal vascular distribution. E A 32-year-old woman at 32 weeks of gestation. The white arrow indicates loss of hypoechoic retroplacental space. The yellow arrow points to hypoechoic or anechoic placental lacunae. F A 32-year-old woman at 32 weeks of gestation with placenta previa. The white arrow indicates loss of hypoechoic retroplacental space. G A 28-year-old woman at 27 weeks of gestation. The white arrow indicates interruption of the continuous hyperechoic bladder line. H A 35-year-old woman at 29 weeks of gestation with placenta previa. The white arrow shows confluent areas of hypoechoic or anechoic placental lacunae. I A 31-year-old woman at 33 weeks of gestation. The yellow arrow demonstrates increased, clustered vascular signals at the placental base. J A 35-year-old woman at 29 weeks of gestation with placenta previa. The white arrow indicates increased vascularity at the placental base. The yellow arrow shows the presence of anechoic placental lacunae. The red arrow indicates blood sinuses of the cervix. K A 28-year-old woman at 27 weeks of gestation. The white arrow indicates bridging vessels at the placental base. The red arrow shows coalescent patches of cervical blood sinuses. L A 29-year-old woman at 27 weeks of gestation with placenta previa. The white arrow indicates bridging vessels and the yellow arrow shows blood sinuses of the cervix. (PL: Placenta; UA: Umbilical Artery; CX: Uterine Cervix; BL: Bladder) All subjects were examined using a Siemens Aera 1.5T magnetic resonance imaging system. During the examination, the pregnant woman lies in the supine position with her feet entering first, wearing earplugs for sound insulation, and allowed to breathe naturally. An 18-channel abdominal phased array coil was employed, centered around the placenta, and the scan covered the area from the diaphragm to the pubic symphysis. The scanning image locations and sequences are detailed in Supplementary Table 2. MRI image data were read by three radiologists with over ten years of work experience (Dr. H, Dr. W, and Dr. L) under blind conditions (unaware of US findings, patient’s placental pathology and surgical results), according to the 11 commonly used MRI features and the recommended standardized terminology (hereinafter referred to as standardized features) in “SAR and ESUR Joint Consensus Statement for MR Imaging of Placenta Accreta Spectrum Disorders”. These 11 MRI features include 7 “recommended” (at least 80% agreement in favor) features (T2-dark intraplacental bands, Placental/uterine bulge, Loss of retroplacental T2-hypointense line, Myometrial thinning, Bladder wall interruption, Focal exophytic mass, Abnormal vascularization of placental bed) and 4 “uncertain” (i.e., consensus was not reached, with less than 80% agreement) features (Placental heterogeneity, Asymmetric shape/thickening of the placenta, Placental ischemic infarction, Abnormal intraplacental vascularity). Inter-observer agreement for MRI feature interpretation was evaluated using Fleiss’ Kappa statistics, disagreements were resolved through collective discussion to reach consensus. The results demonstrated excellent inter-observer agreement, with Kappa values between 0.741 and 0.880 across all features. Statistical analysis was performed with IBM SPSS Statistics 27.0 (SPSS Inc, Chicago, IL). Qualitative variables were described with frequency and percentage, the χ 2 test, Yates’s correction or Fisher’s was used for comparision between groups according to the theoretical frequency; quantitative variables were described as mean ± standard deviation, compared with the t -test between groups. Logistic regression was used for multivariate analysis of the MRI features with inter-group differences detected by the χ 2 test to identify statistically significant MRI indicators. Given the low frequency of some MRI features, leading to a large estimated standard error, the effect sizes were calculated by the bootstrap method following stepwise selection (1000 resamples were performed to ensure model stability). To compare diagnostic schemes, logistic regression (PAS as outcome; MRI, US scores, risk factors as predictors) generated predictive probabilities for ROC analyses (gold standard: surgical/pathological results). We reported AUC with 95% CI and used the Youden index for optimal cutoffs. Delong test was used for the comparison of AUC. For schemes with unclear AUC differences, NRI and IDI measured model improvements. All tests were two-tailed ( α  = 0.05). NRI (net reclassification improvement index, reflecting the improvement in patient classification by the new model compared with the old model); IDI (integrated discrimination improvement index, reflecting the overall improvement in the predictive probability of the model). A p-value of < 0.05 was considered statistically significant.

This study identifies three key standardized MRI features (T2-dark intraplacental bands, Loss of retroplacental T2-hypointense line, Asymmetric shape/thickening of the placenta) and three key risk factors (caesarean delivery history, placenta previa, advanced maternal age) to simplify clinical application. For high-risk pregnant women, a multimodal approach combining risk assessment, ultrasound, and standardized MRI features produced the highest diagnostic performance, but prospective external validation is required before broader clinical implementation.

This study identified three standardized MRI features and three risk factors with significant differences between PAS and control groups. Clinicians can prioritize these high-discriminative features to simplify diagnosis. Moreover, integrating risk factors, US scoring, and these MRI features into a multimodal framework overcomes single-modality limitations. This confirms the combined strategy is most effective, providing a theoretical basis for optimizing PAS diagnostic protocols. Using risk factors alone for prenatal PAS diagnosis yielded an AUC of 0.835, confirming the value of this traditional approach. Compared with US and MRI, it is more cost-effective and convenient, making it suitable for screening high-risk groups. Any factor causing endometrial damage can induce PAS. Injurious procedures (such as CD, myomectomy, uterine curettage, and in vitro fertilization (IVF)) trigger decidual-myometrial defects, which disrupt the basal plate and expose spiral arteries to trophoblast cells, leading to placental invasion [ 14 ]. In uterine scar areas, deficient decidua and damaged myometrium allow deeper extravillous trophoblast invasion. According to FIGO consensus guidelines on PAS, most of the 17 risk factors (5 categories) correlate with direct or indirect uterine damage. Epidemiologically, PAS prevalence rises with CD rates; prior CD history is the most immediate and significant risk factor. Recent research shows that women with CDs have higher risk of PAS. Women with 4 or more CDs face a 9-30-fold higher risk, and those aged ≥ 35 years have elevated risk, likely due to higher CD rates in this group [ 1 ]. A retrospective study involving 68,944 delivery cases showed that the overall CD rate at the research center was as high as 55.0% [ 15 ]. Such a heavy burden of CDs further suggests the potential for a significant increase in the risk of PAS at the population level. In contrast, PAS is rare in patients with uterine malformations, adenomyosis, or submucosal fibroids but with no history of uterine injury, so these are not major risk factors [ 9 ]. IVF embryo transfer is recognized as an independent risk factor [ 16 ], but lacks clear pathophysiological evidence. This study detected significant inter-group differences in three factors: prior CD history, placenta previa, and advanced maternal age. These findings align with current PAS epidemiological characteristics, while no significant differences were observed for other risk factors. MRI is highly accurate for PAS diagnosis, but previous inconsistent descriptions of MRI features hindered inter-professional communication. Standardizing terminology and unifying diagnostic criteria are therefore critical. This study adopted SAR/ESUR joint consensus terms for PAS-related MRI features, to promote their clinical application and academic exchange. Since the 2020 SAR/ESUR consensus proposed standardized PAS MRI features, relevant diagnostic studies have emerged. Thiravit S. et al. [ 17 ] reported three widely used prenatal diagnostic features (the intraplacental low-signal band on T2WI, placenta/uterine bulging, and myometrial thinning), corresponding to the 1st, 2nd, and 4th consensus-recommended features. Romeo V. et al. [ 18 ] confirmed the 1st (the intraplacental low-signal band on T2WI) and 4th (myometrial thinning) features as independently associated with PAS. In this study, 10 of 11 standardized MRI features showed significant inter-group differences. Considering feature frequency variation and single-feature diagnostic limitations, three key features were selected via bootstrap resampling: T2-dark intraplacental bands (1st feature), Loss of retroplacental T2-hypointense line (3rd feature), and Asymmetric shape/thickening of the placenta (9th feature). Consistent with other studies, these three features have high clinical utility; prioritizing their evaluation improves PAS diagnostic efficiency. Their diagnostic values: T2-dark intraplacental bands result from fibrin deposition caused by repeated placental hemorrhage or infarction (sensitivity 0.83, specificity 0.80) [ 19 ]; Placental/uterine bulge is a severe PAS feature (accuracy 90.3%, sensitivity 94.4%, specificity 84.6%) [ 17 ]; Myometrial thinning indicates placental hyperplasia (high sensitivity 97.4%, low specificity 36.4%) [ 8 ]. The accuracy of MRI feature assessment for PAS may be influenced by gestational age, maternal age, examination posture, and the radiologist. As pregnancy progresses, the signal contrast between the placenta and myometrium weakens [ 20 ]. Due to the potential for extreme thinning of myometrium in the third trimester, this feature (Myometrial thinning) should not be used in isolation but rather in conjunction with other features of PAS disorders. Therefore, identifying the most valuable diagnostic markers is one of the key objectives of this study [ 8 ]. Bladder wall interruption (indistinct visualization or traction/compression changes on T2WI) is affected by maternal age, body position, and bladder distension, reducing diagnostic accuracy. Placental tissue in the bladder is a rare but distinctive feature of penetrative PAS [ 8 ], seen only in severe cases. Non-PAS patients may also have asymmetric placental thickening or shape changes (e.g., thickness differences between primary and secondary placentas), requiring multi-angle observation. Risk factors are key for identifying PAS high-risk groups. US and MRI are non-invasive (ionizing radiation-free) assessment methods: US is the first-line prenatal screening tool for its convenience and cost-effectiveness, while MRI provides valuable supplementary information, with the two complementing each other. MRI offers higher diagnostic accuracy for PAS than US, consistent with most studies (e.g., Liu Q et al. [ 21 ]). MRI combined with US also has unique advantages in evaluating posterior uterine wall placental invasion and predicting intraoperative blood loss [ 22 ]. The US sensitivity may be affected by placental location, abdominal wall thickness, and PAS feature typicality, combining with MRI compensates for these limitations to enhance overall efficacy. This study confirmed that MRI + US + risk factors outperforms MRI + US. Maurea S et al. [ 23 ] constructed a similar diagnostic scheme with 91% sensitivity and 88% specificity, matching our combined model’s 91.0% sensitivity but with our specificity slightly higher (97.2%). Sparse data in certain feature categories may have a minor impact on coefficient estimates. Due to variations in ethnicity, regions, and reproductive policies, further clinical validation across diverse populations is needed to confirm the model’s practical applicability. NRI and IDI address limitations of traditional model evaluation by focusing on subtle changes in prediction probability and classification accuracy, enabling more nuanced assessments. These metrics are increasingly applied in high-precision fields like medical diagnosis [ 24 – 26 ]. NRI assesses reclassification improvements, while IDI evaluates overall discriminative enhancement, together providing a comprehensive assessment of model performance. In this study, we used NRI and IDI to compare three diagnostic schemes (MRI+Risk Factors, MRI + US, MRI + US+Risk Factors) where AUC differences were not significant. Although the differences in AUC between the groups were not statistically significant, the NRI suggested improved risk stratification when risk factors were added to MRI + US, indicating incremental value. The main strength of this study is the use of standardized imaging features, which enhance diagnostic reproducibility, reduce inter-radiologist variability, and eliminate terminology ambiguities—enhancing diagnostic traceability and result comparability for clinical and academic use. Additionally, incorporating NRI and IDI to compare multi-method diagnostic efficacy improves the reliability and precision of evaluations, providing valuable references for prenatal PAS diagnosis. The present study has certain limitations. No graded PAS diagnosis (failing to distinguish placenta accreta, placenta increta, and placenta percreta). The significance of relevant MRI features varies among different subtypes: the volume and distribution of T2-dark intraplacental bands increase with the depth of placental invasion, while Loss of retroplacental T2-hypointense line mostly presents as focal interface interruption in accreta and extensive loss accompanied by severe manifestations (e.g., myometrial disruption, bladder invasion) in increta/percreta [ 27 , 28 ]. Single-center data and a study population limited to patients who underwent both US and MRI (introducing potential selection bias and limiting generalizability, though the predominantly high-risk cohort’s homogeneity strengthens applicability to this group); small numbers of severe implantation subtypes; and lack of long-term follow-up (preventing analysis of imaging-outcome associations). Since not all patients underwent hysterectomy, histopathological confirmation was unavailable in some cases. The use of a mixed reference standard may therefore affect diagnostic accuracy estimates. This is common in PAS research and reflects current clinical practice. Radiomics and machine learning offer new opportunities for PAS diagnosis [ 29 ], as radiomics can overcome visual assessment limitations. However, US (the first-line modality) is rarely included in radiomics studies, and existing PAS radiomics studies have heterogeneous methodological quality (generally below ideal levels) [ 7 ], requiring more robust studies. Future research should validate findings in prospective, multi-center, large-sample cohorts. Key directions include developing a PAS severity grading system through further quantitative standardization of MRI features, exploring artificial intelligence and radiomics analysis for objective quantitative diagnosis, and linking optimized diagnostic models to surgical strategies and maternal-fetal outcomes. This study systematically evaluated the diagnostic efficacy of risk factors, ultrasound, and MRI (used alone or in combination) for PAS, and the results confirmed that the combined diagnostic protocol achieved superior diagnostic efficacy. It should be noted, however, that the diagnostic efficacy of both ultrasound and MRI is influenced by multiple factors including team experience and patients’ clinical conditions. In particular, MRI examinations impose high equipment requirements on medical institutions and are associated with increased costs, longer examination times, and stricter demands on patient cooperation. Therefore, even if MRI performs well in tertiary centers, the availability, expertise, and costs of MRI may still vary considerably across different institutions. Given that this study did not include patient populations across different countries and regions, it is recommended that PAS management strategies be flexibly adjusted in light of the actual clinical diagnosis and treatment needs and disparities in resource allocation among different institutions, with the adoption of the following stepwise diagnostic and management algorithm: risk factors combined with ultrasound serves as the initial screening method, and a definitive ultrasound diagnosis can directly guide subsequent clinical management. In cases of an inconclusive ultrasound diagnosis, further complementary MRI examinations are recommended, the results of which can assist in determining the need for surgical intervention or referral to medical institutions with better surgical conditions. For high-risk populations, even if the results of MRI are negative or the diagnosis is inconclusive, adequate preoperative preparation remains essential. This preparation includes establishing a multidisciplinary team and ensuring adequate blood reserve, so as to reduce surgical risks. For medical institutions with limited resources that are unable to perform MRI examinations, active completion of preoperative preparation or timely referral to higher-level medical institutions is advised if PAS is highly suspected clinically, in order to avoid diagnostic and therapeutic delays. In addition, both ultrasound and MRI examinations are highly dependent on the professional experience of operators; systematic professional training or the establishment of a centralized reporting system is therefore required to improve the consistency and reliability of examination results. Overall, the core management strategy for PAS remains based on accurate prenatal diagnosis, with individualized delivery and surgical plans formulated through multidisciplinary collaboration, so as to ultimately reduce the risks of maternal and fetal complications [ 15 , 30 , 31 ].

Placenta accreta spectrum (PAS) is diagnosed at birth when part of the placenta is abnormally attached to the uterine wall. PAS is no longer viewed as a result of aggressive trophoblast invasion but rather as a failure of uterine regulatory mechanisms, with the scar tissue providing a permissive environment for partial abnormal placenta attachment at delivery [ 1 ]. PAS can significantly increase the rate of massive bleeding, hysterectomy and adverse neonatal outcomes [ 2 ], posing a significant risk to both maternal and fetal health. Accurate prenatal diagnosis and intervention are crucial for improving pregnancy outcomes. However, it has been reported that 1/2 to 2/3 of cases fail to receive a definitive diagnosis before delivery, missing the optimal intervention opportunity [ 3 ]. Notably, if PAS is managed by a multidisciplinary team with accurate prenatal diagnosis, conservative uterine surgery can be performed to improve prognosis [ 4 ]. With the increasing caesarean delivery (CD) rate in China, the incidence of PAS has risen, highlighting the need for effective diagnosis and unified management practices [ 5 ]. Currently, the diagnosis of PAS primarily relies on risk factors assessment and ultrasound (US). Although US is widely used due to its convenience and low cost, its diagnostic value for uterine wall and parametrial invasion in PAS is limited. MRI is increasingly used as a confirmatory diagnostic tool in tertiary centers, providing comprehensive placental morphology evaluation [ 6 ]. However, there is no unified MRI diagnostic standard for PAS, and diagnostic accuracy depends on radiologists’ experience, which is unfavorable for individualized diagnosis and clinical decision-making. There is a wide heterogeneity in the use of PAS reference standards among the studies included in the systematic review [ 7 ]. Therefore, standardizing PAS evaluation and developing efficient diagnostic plans is essential for improving PAS prognosis. In 2020, the Society of Abdominal Radiology(SAR) and the European Society of Urogenital Radiology (ESUR) published a joint consensus statement on MRI features of PAS, including 7 “recommended” and 4 “uncertain” features. This provided a unified reference for prenatal MRI examinations [ 8 ]. However, the complexity of these features hinders practical application. In this study, we analyzed the diagnostic value of standardized MRI features from the consensus statement to identify the most discriminative features. We also evaluated the diagnostic efficacy of MRI and US used alone or in combination, as well as risk factors, aiming to provide a comprehensive basis for accurate prenatal PAS diagnosis, optimize the clinical diagnostic pathway, and improve patient outcomes.

Supplementary Material 1. Supplementary Material 1.