Comparative efficacy of MRI versus ultrasound in prenatal diagnosis of placenta accreta spectrum disorders: a single-tertiary center study.

MRI demonstrated a higher performance than US in the subgroup of placental accrete cases. Both MRI and US could correctly diagnose placenta percreta.

A total of 131 pregnant women, ranging in age from 23 to 46 years, delivered at 28 + 2 to 40 + 6 weeks’ gestation, with 130 singletons and one twin pregnancy. Among them, 54 (41.2%) were nulliparous, and 77 (58.8%) were multiparous. Supplementary Table  3 shows the demographic characteristics of the study population. Eight (6.1%) women were diagnosed with hypertensive disease in pregnancy, 21 (16.0%) were diagnosed with gestational diabetes mellitus, 12 (9.2%) were diagnosed with hypothyroidism, 21 (16.0%) had undergone IVF-ET, 10 (7.6%) were diagnosed with uterine fibroids or adenomyosis, and 5 (3.8%) were diagnosed with FGR. Of the 131 pregnant women, 55 (42%) had a history of cesarean section, 2 (1.53%) had undergone myomectomy, and 95 (72.5%) had a curettage or hysteroscopy history. A total of 5 (3.8%) women underwent vaginal delivery and 126 (96.2%) were delivered by cesarean section. There were 32 (24.4%) women obtaining abdominal aorta balloon occlusion during the operation, and the mean blood loss volume during peripartum was 500 mL. Several additional management strategies were employed to improve outcomes, including B-lynch suture in 28 (21.4%) cases, Bakri balloon uterine compression or uterine packing in 47 (35.9%) cases, and selective uterine artery embolization in 12 (9.2%) cases. Seven (5.3%) women in the study underwent hysterectomy. The clinical diagnosis was established after delivery (Supplementary Table  3 ). Among the 131 suspected cases of PAS, 61 (46.5%) were found to be normal, 22 (16.8%) were diagnosed with placenta accreta, 44 (33.6%) with placenta increta, and 4 (3.1%) with placenta percreta. Additionally, placental position diagnoses included 13 (9.9%) with normal position, 4 (3.1%) with low-lying placenta, and 114 (87.0%) with placenta previa. MRI demonstrated an ACC, SEN, and SPE of 0.718, 0.543, and 0.918, respectively, for PAS diagnosis, whereas US yielded an ACC, SEN, and SPE of 0.702, 0.571, and 0.853, respectively (Table  1 ). The inter-observer reading correlation between the two modalities is detailed in Table  2 . The ICC for both modalities was moderate, at 0.663 for MRI and 0.6 for US. Discrepancies between MRI and US diagnoses were outlined in Tables  3 and 4 . For the placenta accreta group, MRI showed slightly superior performance compared to US (ACC for MRI and US: 40.9% versus 27.2%, p = 0.023). For the other three groups—normal, increta, and percreta—both modalities demonstrated equivalent diagnostic capabilities. In the placenta increta group, US had a higher ACC than MRI (0.682 versus 0.568). For the accreta group, both modalities had the poorest performance; however, for the percreta group, both modalities achieved perfect diagnostic accuracy (Fig.  1 ). The potential reasons for misdiagnosis on US in the placenta accreta or increta groups may relate to challenging positions, the acquisition plane of US, thick placental walls, or scar formation due to previous surgery (Figs. 2 , 3 , 4 ). For MRI, possible misdiagnosis in the placenta accreta or increta groups may stem from morphological differences between the time of scanning and the exact time of delivery-typically about one month prior to delivery in our institution (Figs. 5 , 6 ). A summary of the detailed reasons for each modality is presented in Table  4 . Table 1 Performance comparison between MRI and US for PAS confirmation in the included samples (N = 131) taking the open surgery or pathological findings or both as the gold standard AUC (95% CI) SEN (95% CI) SPE (95% CI) PPV (95% CI) NPV (95% CI) ACC (95% CI) MRI 0.730 (0.668–0.801) 0.543 (0.427–0.655) 0.918 (0.817–0.967) 0.884 (0.749–0.953) 0.636 (0.532–0.728) 0.718 (0.634–0.788) US 0.712 (0.638–0.786) 0.571 (0.449–0.693) 0.853 (0.745–0.913) 0.816 (0.612–0.858) 0.634 (0.606–0.794) 0.702 (0.620–0.772) AUC = area under the curve, ICC = inter-observer correlation coefficients, ACC = accuracy, SEN = sensitivity, SPE = specificity, PPV = positive predictive value, NPV = negative predictive value, the parenthesis = 95% confidence interval (CI) Table 2 Inter-observer reading correlation evaluation between two modalities using the randomly selected twenty samples (N = 20) AUC (95% CI) SEN (95% CI) SPE (95% CI) PPV (95% CI) NPV (95% CI) ACC (95% CI) MRI  Reader 1 0.808 (0.633–0.955) 0.889 (0.517–0.997) 0.727 (0.391–0.924) 0.727 (0.391–0.924) 0.889 (0.517–0.997) 0.8 (0.563–0.943)  Reader 2 0.788 (0.599–0.950) 0.667 (0.299–0.925) 0.909 (0.587–0.998) 0.857 (0.487–0.997) 0.769 (0.462–0.950) 0.8 (0.563–0.943) ICC = 0.663, (95 CI% 0.323–0.852), P < 0.001 US  Reader 1 0.899 (0.750–1.000) 0.889 (0.518–0.997) 0.909 (0.587–0.998) 0.889 (0.518–0.997) 0.909 (0.587–0.998) 0.9 (0.683–0.988)  Reader 2 0.788 (0.600–0.950) 0.667 (0.299–0.925) 0.909 (0.587–0.998) 0.857 (0.487–0.997) 0.769 (0.462–0.950) 0.8 (0.563–0.943) ICC = 0.600, (95% CI 0.234–0.819), P = 0.002 AUC  area under the curve, ICC  inter-observer correlation coefficients, ACC accuracy, SEN  sensitivity, SPE  specificity, PPV  positive predictive value, NPV  negative predictive value Table 3 Diagnostic performance comparison between MRI and US across the four subgroups for PAS confirmation in the included samples (N = 131) MRI (right) MRI (wrong) Absence (N = 61) P = 0.481  US (right) 49 3  US (wrong) 8 1 Accrete (N = 22) P = 0.023  US (right) 5 1  US (wrong) 4 12 Increta (N = 44) P = 0.745  US (right) 18 12  US (wrong) 7 7 Percreta (N = 4) NA  US (right) 4 0  US (wrong) 0 0 MRI magnetic resonance imaging, US ultrasound Table 4 Summaries of the discrepancies between MRI and US in PAS misdiagnosing cases (N = 36) N Reasons MRI was right while US was wrong Placenta accreta (N = 7) Placenta increta (N = 4) Normal (N = 8) The possible misdiagnosing explanation on US for placenta accreta or increta may be position challenging for US acquisition (the lower segment of posterior uterine wall(N = 5), the upper segment of uterine (N = 2), the left wall of uterus (N = 1) The possible misdiagnosing explanation on US for normal placenta confirmation: five cases had C-scar surgery history (N = 5) and six cases had thick placenta (N = 6) compromising the placenta display MRI was wrong while US was right Placenta accreta (N = 12) Placenta increta (N = 1) Normal (N = 3) The possible misdiagnosing explanation on MRI for placenta accreta or increta may be the uterine wall edema (N = 2), edema besides cervical internal os (N = 1), focal placental thickness (N = 1) and the early MRI scanning time (approximately one month prior to delivery) (N = 9) The possible misdiagnosing explanation on MRI for normal placenta confirmation (N = 3): lacking experiences Fig. 1 A 35-year-old pregnant woman diagnosed with complete placenta previa at 32 GW. Both US (left) and MR sagittal T2WI (right) showed percreta placental interface (arrowhead). US reveals a placental bulge, thinning of the myometrium, loss of the normal retroplacental space, and interruption of the bladder wall.MR findings show an indistinct boundary between the lower edge of the placenta and the anterior wall of the cervix, the lower anterior wall of the uterus, reaching the serosal surface, and bulging outward towards the bladder. The pregnancy was terminated by cesarean section at 34.2 weeks and placental tissue penetrates the serosal layer of the uterus accompanied by involvement of the bladder, which was confirmed during surgery Fig. 2 A 30-year-old pregnant woman with complete placenta previa. US (left) at 31.5 GW showed suspicious accrete interface (arrowhead) and MR sagittal T2WI (right) at 32.1 GW showed normal interface (arrowhead). US findings show loss of the normal retroplacental space, increased vascular perfusion between the uteroplacental interface reaching the uterine serosa.MR findings show that the boundary between part of the placenta and the myometrium of the uterine sidewall is still clear. The pregnancy was terminated by cesarean section at 37 GW, with no placental adhesion observed Fig. 3 A 38-year-old pregnant woman with marginal placenta previa. US (left) at 35.4 GW showed normal (arrowhead) and MR sagittal T2WI (right) at 29.5 GW showed suspicious accrete interface (arrowhead). US findings show that the normal retroplacental space is continuous.MR findings show that the edges of the placenta are indistinct, with some parts of the placenta having unclear boundaries with the uterine muscle layer. The patient delivered vaginally at 40 GW, with local placental adhesion found postpartum Fig. 4 A 31-year-old pregnant woman with history of myomectomy while the placental position was normal. US (left) at 29.1 GW showed normal interface (arrowhead) and MR sagittal T2WI (right) at 28 GW showed suspicious intreta interface (arrowhead). US indicates that the normal retroplacental space is continuous, and the blood flow signals at the base of the placenta are regular.MR shows a bulging placenta, with indistinct boundaries between the placenta and the left lower segment of the anterior uterine wall muscle layer. The pregnancy was terminated by cesarean section at 39.1 weeks during which regional placental implantation was identified Fig. 5 A 27-year-old pregnant woman with complete placenta previa. US (left) at 33.5 GW showed normal interface (arrowhead) and MR sagittal T2WI (right) at 32.5 GW showed suspicious accrete interface (arrowhead). US findings show loss of the normal retroplacental space. Increased vascular perfusion between the uteroplacental interface reaching the uterine serosa are seen.MR findings show no obvious indistinct boundaries between the placenta and the uterine muscle layer. The pregnancy was terminated by cesarean section at 34 weeks, with no placental adhesion observed Fig. 6 A 24-year-old pregnant woman with complete placenta previa. US (left) at 29.1 GW showed suspicious intreta interface (arrowhead) and MR sagittal T2WI (right) at 28 GW showed normal interface (arrowhead). US findings indicate loss of the normal retroplacental space, and increased vascular perfusion are observed.MR findings show that the boundary between the placenta and the uterine muscle layer is still discernible. The pregnancy was terminated by cesarean section at 33.4 weeks during which placental implantation was identified Performance comparison between MRI and US for PAS confirmation in the included samples (N = 131) taking the open surgery or pathological findings or both as the gold standard AUC = area under the curve, ICC = inter-observer correlation coefficients, ACC = accuracy, SEN = sensitivity, SPE = specificity, PPV = positive predictive value, NPV = negative predictive value, the parenthesis = 95% confidence interval (CI) Inter-observer reading correlation evaluation between two modalities using the randomly selected twenty samples (N = 20) AUC  area under the curve, ICC  inter-observer correlation coefficients, ACC accuracy, SEN  sensitivity, SPE  specificity, PPV  positive predictive value, NPV  negative predictive value Diagnostic performance comparison between MRI and US across the four subgroups for PAS confirmation in the included samples (N = 131) MRI magnetic resonance imaging, US ultrasound Summaries of the discrepancies between MRI and US in PAS misdiagnosing cases (N = 36) Placenta accreta (N = 7) Placenta increta (N = 4) Normal (N = 8) The possible misdiagnosing explanation on US for placenta accreta or increta may be position challenging for US acquisition (the lower segment of posterior uterine wall(N = 5), the upper segment of uterine (N = 2), the left wall of uterus (N = 1) The possible misdiagnosing explanation on US for normal placenta confirmation: five cases had C-scar surgery history (N = 5) and six cases had thick placenta (N = 6) compromising the placenta display Placenta accreta (N = 12) Placenta increta (N = 1) Normal (N = 3) The possible misdiagnosing explanation on MRI for placenta accreta or increta may be the uterine wall edema (N = 2), edema besides cervical internal os (N = 1), focal placental thickness (N = 1) and the early MRI scanning time (approximately one month prior to delivery) (N = 9) The possible misdiagnosing explanation on MRI for normal placenta confirmation (N = 3): lacking experiences A 35-year-old pregnant woman diagnosed with complete placenta previa at 32 GW. Both US (left) and MR sagittal T2WI (right) showed percreta placental interface (arrowhead). US reveals a placental bulge, thinning of the myometrium, loss of the normal retroplacental space, and interruption of the bladder wall.MR findings show an indistinct boundary between the lower edge of the placenta and the anterior wall of the cervix, the lower anterior wall of the uterus, reaching the serosal surface, and bulging outward towards the bladder. The pregnancy was terminated by cesarean section at 34.2 weeks and placental tissue penetrates the serosal layer of the uterus accompanied by involvement of the bladder, which was confirmed during surgery A 30-year-old pregnant woman with complete placenta previa. US (left) at 31.5 GW showed suspicious accrete interface (arrowhead) and MR sagittal T2WI (right) at 32.1 GW showed normal interface (arrowhead). US findings show loss of the normal retroplacental space, increased vascular perfusion between the uteroplacental interface reaching the uterine serosa.MR findings show that the boundary between part of the placenta and the myometrium of the uterine sidewall is still clear. The pregnancy was terminated by cesarean section at 37 GW, with no placental adhesion observed A 38-year-old pregnant woman with marginal placenta previa. US (left) at 35.4 GW showed normal (arrowhead) and MR sagittal T2WI (right) at 29.5 GW showed suspicious accrete interface (arrowhead). US findings show that the normal retroplacental space is continuous.MR findings show that the edges of the placenta are indistinct, with some parts of the placenta having unclear boundaries with the uterine muscle layer. The patient delivered vaginally at 40 GW, with local placental adhesion found postpartum A 31-year-old pregnant woman with history of myomectomy while the placental position was normal. US (left) at 29.1 GW showed normal interface (arrowhead) and MR sagittal T2WI (right) at 28 GW showed suspicious intreta interface (arrowhead). US indicates that the normal retroplacental space is continuous, and the blood flow signals at the base of the placenta are regular.MR shows a bulging placenta, with indistinct boundaries between the placenta and the left lower segment of the anterior uterine wall muscle layer. The pregnancy was terminated by cesarean section at 39.1 weeks during which regional placental implantation was identified A 27-year-old pregnant woman with complete placenta previa. US (left) at 33.5 GW showed normal interface (arrowhead) and MR sagittal T2WI (right) at 32.5 GW showed suspicious accrete interface (arrowhead). US findings show loss of the normal retroplacental space. Increased vascular perfusion between the uteroplacental interface reaching the uterine serosa are seen.MR findings show no obvious indistinct boundaries between the placenta and the uterine muscle layer. The pregnancy was terminated by cesarean section at 34 weeks, with no placental adhesion observed A 24-year-old pregnant woman with complete placenta previa. US (left) at 29.1 GW showed suspicious intreta interface (arrowhead) and MR sagittal T2WI (right) at 28 GW showed normal interface (arrowhead). US findings indicate loss of the normal retroplacental space, and increased vascular perfusion are observed.MR findings show that the boundary between the placenta and the uterine muscle layer is still discernible. The pregnancy was terminated by cesarean section at 33.4 weeks during which placental implantation was identified

This study conducted a retrospective analysis of clinical data from patients suspected of having Placenta Accreta Spectrum (PAS) at the Obstetrics and Gynecology Hospital of Fudan University, spanning the period from January 2020 to December 2023. The inclusion criteria were as follows: (1) aged 18 years or older; (2) having had more than three pregnancies, including the current one, or having undergone at least one prior uterine procedure (such as cesarean section, myomectomy, curettage, or hysteroscopy), or being diagnosed with placenta previa or a low-lying placenta; (3) having undergone both ultrasonographic and MRI examinations due to suspicion of PAS. The exclusion criteria included the presence of severe fetal abnormalities, loss of follow-up, and cases where the patient could not be evaluated using the ultrasonic scoring system or MRI. Ethical approval for this study was granted by our Institutional Ethics Committee. We employed an ultrasonic scoring system to predict and assess PAS [ 5 ] (Supplementary Table  1 ) after 28 weeks of gestation and repeated it between 34 and 36 weeks if available. The two sonographers who performed the scans had more than ten years of professional experience and at least five years of specialized work with the ultrasonic scoring system. The ultrasound equipment utilized in our institution included a Philips IU22 and a GE Voluson E8 color Doppler ultrasound diagnostic device, both equipped with a 3.5 MHz probe. Patients were routinely positioned supine, with the option to be placed in a lateral position if necessary, and were instructed to have their bladder filled to approximately 300 ml before the examination. The placenta was scanned in longitudinal, transverse, and coronal sections. Prenatal ultrasonographic features including position and thickness of the placenta, loss of the normal retroplacental space, bladder wall interruption, existence of abnormal placental lacunae, increased vascular perfusion between the uteroplacental interface reaching the uterine serosa, completeness of cervical morphology, existence of cervical sinus, and history of cesarean section. Each feature was assigned a score of 0, 1, or 2, and a cumulative score was calculated for each patient. Placenta accreta and its severe forms, such as placenta increta and percreta, can be distinguished using a cut-off score of ≥ 5, while a score of ≥ 10 indicates a higher risk of placenta percreta. Two sonography physicians, each with more than 10 years of experience, conducted the aforementioned PAS diagnosis. In our hospital, an MRI examination was scheduled after 28 weeks of gestation and no later than 36 weeks. Repeat examinations were generally not performed. MR imaging was conducted using a 1.5-T MR system (Magnetom Avanto, Siemens) and a 3.0-T MR system (Ingenia 3.0T CX, Philips Medical Systems, Amsterdam, the Netherlands). All MRI examinations were performed without sedation. The routine 1.5-T MRI protocols for fetal assessment included axial, sagittal, and coronal T2-weighted imaging (T2WI) with half-Fourier acquisition single-shot turbo spin echo (HASTE) and axial, sagittal, and coronal true fast imaging with steady-state precession (True-FISP) sequences. The routine 3.0-T MRI protocols included axial, sagittal, and coronal T2-weighted imaging (T2WI) with single-shot turbo spin echo (SSTSE) and axial, sagittal, and coronal true fast imaging with balanced fast field echo (BFFE) sequences (Supplementary Table  2 ). For all MRI sequences, the specific absorption rate was controlled under 2.0 W/kg. The MRI features indicative of PAS, as described in the literature, primarily included the interface contour (clear or ambiguous) between the placenta and myometrium, T2 dark bands in the placenta, focal placental or uterine bulge, thickened placenta, the presence of abnormally prominent vessels within the uteroplacental interface, and the clarity of the discrimination between the bladder and uterine wall. All readings were performed independently and were blinded to the final pathological diagnosis on the picture archiving and communication system (PACS) server by two readers, each with more than 10 years of experience in obstetric imaging. Two readers, one with more than 10 years and the other with more than 6 years of experience in obstetric imaging, interpreted the MRI findings separately. In the event of a disagreement, the senior reader’s conclusion was considered the final diagnosis. All readers were aware of the project and understood that the MRI data originated from patients suspected of having PAS. Positive PAS findings on MRI were graded on a 5-point scale, reflecting the diagnostic confidence in detecting PAS: 5 = definitely present, 4 = possibly present, 3 = equivocal, 2 = possibly absent, and 1 = definitely absent. The collected data included maternal age, gravidity, parity, maternal comorbidities (such as hypertensive disease in pregnancy, gestational diabetes mellitus, hypothyroidism, in vitro fertilization and embryo transfer [IVF-ET], uterine fibroids or adenomyosis, fetal growth restriction [FGR], twin pregnancies), history of uterine operations (including cesarean section, myomectomy, and curettage or hysteroscopy), gestational of delivery, mode of delivery, apply abdominal aorta balloon occlusion or not, blood loss volume during peripartum, and extra management intra- and postpartum (including B-lynch suture, Bakri balloon uterine compression or uterine packing, selective uterine artery embolism, and hysterectomy). The diagnosis of placenta position (normal, low-lying, or previa) and placenta invasion (normal, accrete, increta, or percreta) were also recorded. The diagnosis of placenta invasion was determined based on clinical outcomes. Experienced obstetricians were responsible for conducting evaluations and establishing diagnoses during cesarean section or post-vaginal delivery. Pertaining to PAS, these conditions were delineated based on both the patient’s clinical and histopathological assessments. Clinically, PAS was ascertained as an unsuccessful endeavor to extricate the placenta during cesarean section. Histologically, PAS was characterized sections showing extended areas of absent decidua between villous tissue and myometrium. The diagnosis of PAS after excision is based on microscopic examination of the placental bed [ 2 , 5 ]. Clinical diagnoses categorized patients into four groups [ 10 ]: Absence of PAS: During cesarean section or vaginal delivery, the placenta was completely expelled during the third stage of labor, with normal placental adhesion. Placenta accreta: Placenta accreta occurs when chorionic villi attach directly to the myometrium. During cesarean section, the placental tissue did not invade the serosal layer of the uterus. Attempts at manual placental separation results in heavy bleeding [ 11 ]. Placenta increta: Placenta increta occurs when chorionic villi invade into the myometrium. During cesarean section, the placental tissue did not invade the serosal layer of the uterus. Despite the use of uterotonic drugs and gentle traction on the umbilical cord, the placenta did not separate, requiring mechanical or surgical removal of the placenta. The placenta was fully adherent. During vaginal delivery, manual removal of the placenta was necessary due to complete adhesion of the placenta, and it was unsuccessful due to abnormal placental invasion [ 11 ]. Placenta percreta: Placenta percreta occurs when chorionic villi invade through the myometrium and serosa and sometimes into adjacent organs. During cesarean section, placental tissue penetrates the serosal layer of the uterus, which may be accompanied by the involvement of the bladder or other pelvic tissues or organs [ 11 ]. Absence of PAS: During cesarean section or vaginal delivery, the placenta was completely expelled during the third stage of labor, with normal placental adhesion. Placenta accreta: Placenta accreta occurs when chorionic villi attach directly to the myometrium. During cesarean section, the placental tissue did not invade the serosal layer of the uterus. Attempts at manual placental separation results in heavy bleeding [ 11 ]. Placenta increta: Placenta increta occurs when chorionic villi invade into the myometrium. During cesarean section, the placental tissue did not invade the serosal layer of the uterus. Despite the use of uterotonic drugs and gentle traction on the umbilical cord, the placenta did not separate, requiring mechanical or surgical removal of the placenta. The placenta was fully adherent. During vaginal delivery, manual removal of the placenta was necessary due to complete adhesion of the placenta, and it was unsuccessful due to abnormal placental invasion [ 11 ]. Placenta percreta: Placenta percreta occurs when chorionic villi invade through the myometrium and serosa and sometimes into adjacent organs. During cesarean section, placental tissue penetrates the serosal layer of the uterus, which may be accompanied by the involvement of the bladder or other pelvic tissues or organs [ 11 ]. Continuous variables were presented as the median ± standard deviation and range and were compared with Student’s t test. Categorical variables were reported as percentages and were compared with the chi-square test and Fisher’s exact test. All analyses were conducted using SPSS (version 23.0, IBM). The diagnostic performance of the Placenta Accreta Spectrum (PAS) was evaluated against the gold standard of pathological diagnosis, assessing accuracy (ACC), sensitivity (SEN), specificity (SPE), positive predictive value (PPV), and negative predictive value (NPV). A P-value of less than 0.05 was considered to indicate statistical significance. Inter-observer reliability was evaluated using the intraclass correlation coefficient (ICC) (moderate: 0.40–0.59; good: 0.60–0.79; excellent: ≥ 0.80) for each modality, assessed in a random sample of twenty pregnancies. Two reviewers, blinded to the final outcomes, independently interpreted the imaging data.

This study presented a comparative analysis of the diagnostic capabilities of MRI and US in the confirmation of PAS. Our results underscore that both MRI and US have equivalent overall accuracy rates, with MRI achieving an accuracy of 0.718 and US an accuracy of 0.702. However, a notable discrepancy emerged in the detection of placenta accreta, where MRI and US accuracy rates were 0.409 and 0.272, respectively (p = 0.023), indicating a need for enhanced diagnostic precision in this area. Conversely, both modalities excelled in the accurate depiction of placenta percreta, a finding that consolidates their clinical utility in such cases. PAS characterizes a clinical scenario in which the placenta fails to separate spontaneously postpartum. Affected patients are at an elevated risk for significant hemorrhage, which not only necessitates substantial blood transfusion requirements but also poses a life-threatening risk to the parturient. Furthermore, PAS escalates the likelihood of hysterectomy, which in turn can precipitate a cascade of adverse outcomes. Optimal maternal and neonatal outcomes are more commonly achieved when a prenatal diagnosis of PAS is established, enabling the patient to be overseen by a multidisciplinary team (MDT) with specialized knowledge in managing this condition. At our institution, an MDT for PAS is spearheaded by the Department of Obstetrics, with collaborative involvement from the Departments of Gynecology, Neonatology, Anesthesiology, Radiology, Urology, Nursing, and the Blood Bank, all of which contribute their respective expertise to the comprehensive care of patients with PAS [ 12 ]. The team also set up an MDT clinic to assess the overall maternal and fetal condition starting from the late second trimester or early third trimester. This clinic is dedicated to discussing the preparations necessary for the perinatal period and devising a personalized delivery plan tailored to each patient’s unique circumstances [ 13 ]. In the past, the antepartum evaluation of PAS disorders relied predominantly on clinical experience and practice. Key factors identified as major risk factors included advanced maternal age, a history of cesarean section, previous uterine surgery, and the presence of placenta previa during pregnancy. In recent years, an ultrasonic scoring system has been implemented to prognosticate the type and severity of PAS. Additionally, MRI has been utilized to ascertain the placental location and PAS classification [ 14 ]. The ultrasonic scoring system assigns scores to various parameters, including the location of the placenta, the thickness of the placenta in the lower segment of the uterus, the retroplacental hypoechoic, the bladder line, the placental lacunae, the blood flow signal at the base of the placenta, the cervical sinus, the shape of the cervix, and the number of cesarean sections experienced by the patient [ 15 ]. MRI offers high-resolution imaging of soft tissues and is advantageous as it is unaffected by fetal position, placental location, or maternal body habitus [ 16 ]. Consequently, MRI is employed to determine placental positioning, the nature of placental implantation, and the diameter of the abdominal aorta. This information is instrumental in the preoperative selection of the appropriate catheter model for abdominal aortic balloon placement. Clinical history, ultrasound, and MRI examination serve as the foundation for evaluating the risk of placenta accreta. These assessments inform critical preoperative decisions, including the potential placement of abdominal aortic balloons and ureteral stents, the selection of uterine incision sites, and the consideration of conservative surgical approaches for delivery management. While ultrasound and MRI are prevalent in clinical practice, with MRI being more frequently found in academic tertiary centers, ultrasound remains a mainstay for diagnosing placenta previa. The effectiveness of ultrasound in PAS diagnosis is limited by the absence of a definitive sign or sign combination that accurately quantifies the degree of placental invasion. Furthermore, diagnostic accuracy is contingent upon the operator’s technical proficiency and subjective assessment [ 7 ]. Additional variables such as bladder distention, abdominal wall adiposity, and bowel gas patterns can also influence the accuracy of PAS diagnosis. In this study, the ultrasonic scoring system encompassed both imaging characteristics of PAS and pertinent clinical data, including placental location and history of cesarean sections, which are significant risk factors. Despite these comprehensive criteria, challenges in detecting placental positions with the ultrasound probe and discerning local placental thickening and myometrial morphological changes can lead to misdiagnosis [ 17 , 18 ]. MRI is advocated as a second-line diagnostic modality for PAS, offering a detailed assessment of the depth and lateral extension of myometrial invasion, which is particularly valuable in cases of posterior placentation and when US suggests parametrial invasion. MRI facilitates subsequent re-evaluation by multiple physicians, enhancing diagnostic accuracy. Despite its higher cost and limited availability compared to US, it remains a crucial adjunct in complex cases. In instances where PAS pregnancies remain stable in the third trimester, our institution contemplates the periodic re-evaluation of ultrasonic scoring, typically at one-month intervals. It should be acknowledged that the prospect of multiple MRI examinations during pregnancy can be daunting for many women, often due to economic considerations or the discomfort associated with the noise generated by the MRI machinery. This hesitancy is corroborated by our study’s findings, which indicate that early MRI scanning may potentially lead to missed diagnoses. Compared with the results reported in the literature, our findings did not achieve the high performance levels expected for each modality [ 19 ]. Several potential reasons may account for this discrepancy. Firstly, our study encompassed all clinical scenarios associated with PAS, and in some subgroups, both modalities demonstrated relatively low performance, particularly for the detection of accreta placenta. However, given the high SPE values, both modalities could effectively rule out negative cases in advance. Secondly, the inter-observer interpretation consistency in this study was only moderate, indicating that the readers’ experiences or habits significantly influenced the final diagnosis. Although key MRI features for PAS diagnosis have been well-documented, in some cases—especially within the accreta group—diagnostic confidence remained suboptimal or incorrect [ 20 , 21 ]. There was also literature reporting the use of clinical and imaging data, or a combination of both, to predict the risk of PAS, which has demonstrated useful clinical value [ 22 ]. Furthermore, scar dehiscence has been identified as a contributor to false positive PAS diagnoses [ 23 ]. The current scoring systems appear inadequate in distinguishing true PAS cases from non-PAS scar dehiscence cases [ 24 ]. In our previous studies, we attempted to develop a machine-learning model to assist clinicians in assessing the potential for placental accreta [ 25 , 26 ]. In the future, prenatal diagnosis of PAS disorders will require dynamic assessment based on MDT diagnosis and treatment, supplemented by artificial intelligence when necessary, which can also aid in timely disease referral and reduce maternal and neonatal complications. In addition, in recent years, there has been increasing consensus amongst PAS experts that the location of the PAS lesion correlates with the morbidity risk and outcome rather than the depth of invasion [ 27 , 28 ]. It will also be investigated in future studies. There are also some limitations in our work. Firstly, although the dataset size in this study was comparable to or larger than those in previous studies, it was still relatively small and derived from a single center. Secondly, the current clinical criteria for PAS confirmation in clinical practice are based on postpartum findings, including clinical features and histological results. Due to the retrospective nature of the study, only pregnancies followed until delivery were included, which may have led to an overestimation of the true results. Thirdly, as mentioned earlier, the readers’ experiences certainly influenced the decision-making process. In the present study, the inter-observer correlation coefficients for both modalities were moderate, introducing a potential observer bias. Additionally, the use of various magnetic modalities may also have influenced the final PAS diagnosis [ 29 ]. In conclusion, MRI provided additional value for pregnancies suspected of placental accrete, and US demonstrated better ACC than MRI in depicting increta placenta. Overall, MRI and US exhibited equivalent performance in confirming PAS.

Placenta accreta spectrum (PAS) disorders are characterized by abnormal placental adherence to the myometrium rather than the uterine decidua. These conditions are classified based on the depth of placental villi invasion: placenta accreta, where the placenta attaches directly to the myometrium without interposed decidua; placenta increta, with invasion into the myometrium; and placenta percreta, which extends through the serosa and may involve adjacent structures [ 1 , 2 ]. PAS has been associated with an increased risk of peripartum and postpartum hemorrhage (PPH), necessitating hysterectomy, and the potential for other organ injuries [ 3 ]. Newborns from pregnancies affected by placenta accreta spectrum disorders are at an increased risk for negative neonatal outcomes, including low Apgar scores [ 4 ]. Prenatal screening is instrumental in identifying women suspected of having PAS, facilitating appropriate delivery planning, and contributing to improved clinical outcomes. Both Ultrasound (US) and Magnetic Resonance Imaging (MRI) are utilized for prenatal screening and diagnosis of PAS, demonstrating high sensitivity and specificity [ 5 , 6 ]. US is often the preferred initial imaging modality due to its lower cost and widespread availability. The accuracy of US for diagnosing placenta accreta is reported to be high, although it is dependent on the operator’s experience [ 7 ]. MRI is reserved for cases where US findings are inconclusive or when additional diagnostic clarity is required for suspected PAS [ 8 ]. Multiplanar MRI images will help to image both the placenta and fetus at a one-time scan. It was also reported that some key findings on MRI enhance the accuracy of PAS diagnosis [ 9 ]. Given that pathological confirmation is not always feasible for suspected PAS and that the final diagnosis often relies on clinical outcomes, further clarification on the comparative diagnostic performance of US and MRI is warranted. Our study, conducted at a single institution, focuses on clinical outcomes as the primary endpoint. The aim is to compare the diagnostic performance of US and MRI in confirming PAS and to identify the optimal imaging modality to inform individualized treatment strategies.

Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 21 KB) Supplementary file1 (DOCX 21 KB)