Diagnostic accuracy of magnetic resonance imaging in assessing placental adhesion disorder in patients with placenta previa: Correlation with histological findings

Magnetic resonance imaging of abnormal placental shapes with correlation with pathologic findings

OBJECTIVE

This study aimed to investigate the magnetic resonance imaging (MRI) characteristics of abnormal placental shapes (APS) compared with prenatal ultrasound.

METHODS

From an initial cohort of 613 women with a high prevalence of placenta accreta spectrum (PAS) disorders, the MRI findings of 27 pregnant women with APS who underwent antenatal ultrasound and MRI examinations before delivery were retrospectively analyzed. The clinicopathological findings were used as the gold standard, and the sensitivity, specificity, and accuracy of antenatal MRI and a multidisciplinary team experienced in diagnosing APS were assessed.

RESULTS

The 27 patients diagnosed with APS included 14 cases of succenturiate placenta, eight cases of the bilobed placenta, two cases of the circumvallate placenta, and one case each of placenta chorioangioma, placenta membranacea, and placental mesenchymal dysplasia. The sensitivity and specificity of APS classification with antenatal MRI were 40.74% (11/27) and 97.65% (498/510), respectively. Nonetheless, the multidisciplinary team achieved a higher sensitivity and specificity of up to 96.29% (26/27) and 99.22% (506/510), respectively.

CONCLUSION

We have demonstrated the complementary role of MRI and ultrasound in the detection of placental shapes in the setting of MRI images, highlighting the importance of radiologists communicating with sonographers in the diagnosis of APS.

Prediction of placenta accreta spectrum in patients with placenta previa using a clinical, US and MRI combined model: A retrospective study with external validation

PURPOSE

To build and validate a predictive model of placental accreta spectrum (PAS) in patients with placenta previa (PP) combining clinical risk factors (CRF) with US and MRI signs.

METHOD

Our retrospective study included patients with PP from two institutions. All patients underwent US and MRI examinations for suspicion of PAS. CRF consisting of maternal age, cesarean section number, smoking and hypertension were retrieved. US and MRI signs suggestive of PAS were evaluated. Logistic regression analysis was performed to identify CRF and/or US and MRI signs associated with PAS considering histology as the reference standard. A nomogram was created using significant CRF and imaging signs at multivariate analysis, and its diagnostic accuracy was measured using the area under the binomial ROC curve (AUC), and the cut-off point was determined by Youden's J statistic.

RESULTS

A total of 171 patients were enrolled from two institutions. Independent predictors of PAS included in the nomogram were: 1) smoking and number of previous CS among CRF; 2) loss of the retroplacental clear space at US; 3) intraplacental dark bands, focal interruption of the myometrial border and placental bulging at MRI. A PAS-prediction nomogram was built including these parameters and an optimal cut-off of 14.5 points was identified, showing the highest sensitivity (91%) and specificity (88%) with an AUC value of 0.95 (AUC of 0.80 in the external validation cohort).

CONCLUSION

A nomogram-based model combining CRF with US and MRI signs might help to predict PAS in PP patients, with MRI contributing more than US as imaging evaluation.

Segmentation methods applied to MRI-derived radiomic analysis for the prediction of placenta accreta spectrum in patients with placenta previa

PURPOSE

To retrospectively evaluate the performance of different manual segmentation methods of placenta MR images for predicting Placenta Accreta Spectrum (PAS) disorders in patients with placenta previa (PP) using a Machine Learning (ML) Radiomics analysis.

METHODS

64 patients (n=41 with PAS and n= 23 without PAS) with PP who underwent MRI examination for suspicion of PAS were retrospectively selected. All MRI examinations were acquired on a 1.5 T using T2-weighted (T2w) sequences on axial, sagittal and coronal planes. Ten different manual segmentation methods were performed on sagittal placental T2-weighted images obtaining five sets of 2D regions of interest (ROIs) and five sets of 3D volumes of interest (VOIs) from each patient. In detail, ROIs and VOIs were positioned on the following areas: placental tissue, retroplacental myometrium, cervix, placenta with underneath myometrium, placenta with underneath myometrium and cervix. For feature stability testing, the same process was repeated on 30 randomly selected placental MRI examinations by two additional radiologists, working independently and blinded to the original segmentation. Radiomic features were extracted from all available ROIs and VOIs. 100 iterations of 5-fold cross-validation with nested feature selection, based on recursive feature elimination, were subsequently run on each ROI/VOI to identify the best-performing method to classify instances correctly.

RESULTS

Among the segmentation methods, the best performance in predicting PAS was obtained by the VOIs covering the retroplacental myometrium (Mean validation score: 0.761, standard deviation: 0.116).

CONCLUSION

Our preliminary results show that the VOI including the retroplacental myometrium using T2w images seems to be the best method when segmenting images for the development of ML radiomics predictive models to identify PAS in patients with PP.

Imaging the Uterus in Placenta Accreta Spectrum Disorder

Antenatal diagnosis of placenta accreta spectrum (PAS) improves maternal and neonatal outcomes by allowing for multidisciplinary planning and preparedness. Ultrasound is the primary imaging tool. Simplification and standardization of placental evaluation and reporting terminology allows improved communication and understanding between teams. Prior to 10 weeks of gestation, gestational sac position and least myometrial thickness surrounding the gestational sac help PAS diagnosis very early in pregnancy. Late first-, second-, and third-trimester evaluation includes comprehensive evaluation of the placenta, transabdominal and transvaginal with partially full maternal urinary bladder, and by color Doppler. Subsequently, the sonologist should indicate whether the evaluation was optimal or suboptimal; the level of suspicion as low, moderate, or high; and the extent as focal, global, or extending beyond the uterus. Other complementary imaging modalities such as 3D-power Doppler ultrasound, magnetic resonance imaging (MRI), and vascular topography mapping strive to improve antenatal placental evaluation but remain investigational at present. KEY POINTS: · Antenatal imaging, primarily using ultrasound with partially full maternal urinary bladder, is an essential means of evaluation of those at risk for PAS.. · Simplification and standardization of placental evaluation and reporting will allow improved communication between the multidisciplinary teams.. · Gestational sac location prior to 10 weeks of gestation and four markers after that (placental lacunae and echostructure, myometrial thinning, hypoechoic zone with or without bulging between placenta and myometrium, and increased flow on color Doppler)..

Bladder involvement in placenta accreta spectrum disorder with placenta previa: MRI findings and outcomes correlation

BACKGROUND

Placental accreta spectrum (PAS) disorder with bladder involvement can be associated with maternal and neonatal morbidity. Magnetic resonance imaging (MRI) may provide accurate preoperative diagnoses.

OBJECTIVE

This study had 2 aims: to retrospectively review the MRI findings for bladder involvement in PAS with placental previa and to correlate bladder involvement with maternal and neonatal outcomes.

MATERIALS AND METHODS

MRI images of 48 patients with severe PAS (increta and percreta) with placenta previa/low-lying placenta were evaluated by 2 experienced radiologists blinded to the final diagnoses. Nine MRI findings and stepwise logistic regression analysis were assessed to identify predictive MRI findings for bladder involvement. The correlations between PAS patients with bladder involvement and clinical outcomes were analyzed using Fisher's exact test.

RESULTS

Of the 48 patients, 27 did not have bladder involvement, while 21 did. Logistic regression analysis identified 2 predictive MRI features for bladder involvement. They were abnormal vascularization (OR,6.94; 95 %CI,1.05-45.75) and loss of the chemical shift line at the uterovesical interface (OR, 4.41; 95 %CI, 0.63-30.98). The sensitivity and specificity of the combined MRI features were 38.1 % and 100 %, respectively (p = 0.001). A significant correlation was found between bladder involvement and massive blood loss during surgery (p = 0.022).

CONCLUSIONS

PAS with bladder involvement was significantly correlated with massive surgical blood loss. Prenatally, the disorder was predicted with high specificity by the combination of loss of chemical shift artifacts in the steady-state free precession sequence and abnormal vascularization at the uterovesical interface on MRI.

A prospective cohort study: can advanced ultrasonography replace magnetic resonance imaging in the diagnosis of placental adhesion disorders?

OBJECTIVES

To define and compare the diagnostic accuracy of ultrasonography (USG) and magnetic resonance imaging (MRI) for the placental adhesive disorder (PAD).

METHODS

A prospective study was conducted between January 2019 and February 2020 in a tertiary referral university hospital. A total of 115 placenta previa cases were included in the study during the third trimester of gestation. USG examination was performed, and the placenta was scanned in a systematic manner using gray-scale ultrasound, color Doppler flow mapping, and 3-D imaging for each participant. Thereafter, all participants underwent an MRI examination. USG and MRI findings were compared with histopathological findings.

RESULTS

Loss of the retroplacental sonolucent zone (71% [95% CI 47-88]) and an irregular retroplacental sonolucent zone (71% [95% CI 47-88]) were the most sensitive USG parameters. For MRI, the uterine bulging parameter was the most sensitive (60% [95% CI 36-80]) and specific (91% [95% CI 83-96]) findings, and it had the highest accuracy rate (85% [95% CI 77-91]). Overall, the USG sensitivity, specificity, and accuracy rates were 77% (95% CI 54-92), 87% (95% CI 79-93), and 85% (95% CI 77-91), respectively. The MRI sensitivity, specificity, and accuracy rates for all participants were 81% (95% CI 59-94), 85% (95% CI 76-92), and 84% (95% CI 76-90), respectively.

CONCLUSIONS

In the diagnosis of PAD, the specificity and accuracy of USG are higher than that of MRI, whereas the sensitivity of MRI is better than that of USG.

Assessment of postpartum haemorrhage for placenta accreta: Is measurement of myometrium thickness and dark intraplacental bands using MRI helpful?

Background

This study aimed to investigate the predictive values of magnetic resonance imaging (MRI) myometrial thickness grading and dark intraplacental band (DIB) volumetry for blood loss in patients with placenta accreta spectrum (PAS).

Methods

Images and clinical data were acquired from patients who underwent placenta MRI examinations and were diagnosed with PAS from March 2015 to January 2021. Two radiologists jointly diagnosed, processed, and analysed the MR images of each patient. The analysis included MRI-based determination of placental attachment, as well as myometrial thickness grading and DIB volumetry. The patients included in the study were divided into three groups according to the estimated blood loss volume: in the general blood loss (GBL) group, the estimated blood loss volume was < 1000 ml; in the massive blood loss (MBL) group, the estimated blood loss volume was ≥ 1000 ml and < 2000 ml; and in the extremely massive blood loss (ex-MBL) group, the estimated blood loss volume was ≥ 2000 ml. The categorical, normally distributed, and non-normally distributed data were respectively analysed by the Chi-square, single-factor analysis of variance, and Kruskal–Wallis tests, respectively. The verification of correlation was completed by Spearman correlation analysis. The evaluation capabilities of indicators were assessed using receiver operating characteristic curves.

Results

Among 75 patients, 25 were included in the GBL group, 26 in the MBL group, and 24 in the ex-MBL group. A significant negative correlation was observed between the grade of myometrial thickness and the estimated blood loss (P < 0.001, ρ = − 0.604). There was a significant positive correlation between the volume of the DIB and the estimated blood loss (P < 0.001, ρ = 0.653). The areas under the receiver operating characteristic curve of the two MRI features for predicting blood loss ≥ 2000 ml were 0.776 and 0.897, respectively.

Conclusions

The grading and volumetric MRI features, myometrial thickness, and volume of DIB, can be used as good prediction indicators of the risk of postpartum haemorrhage in patients with PAS.

Placenta Accreta Spectrum Disorders and Radiomics: Systematic review and quality appraisal

PURPOSE

Ultrasound and magnetic resonance imaging are the imaging modalities of choice for placenta accrete spectrum (PAS) disorders assessment. Radiomics could further increase the value of medical images and allow to overcome the limitations linked to their visual assessment. Aim of this systematic review was to identify and appraise the methodological quality of radiomics studies focused PAS disorders applications.

METHOD

Three online databases (PubMed, Scopus and Web of Science) were searched to identify original research articles on human subjects published in English. For the qualitative synthesis of results, data regarding study design (e.g., retrospective or prospective), purpose, patient population (e.g., sample size), imaging modalities and radiomics pipelines (e.g., segmentation and feature extraction strategy) were collected. The appraisal of methodological quality was performed using the Radiomics Quality Score (RQS).

RESULTS

10 articles were finally included and analyzed. All were retrospective and MRI-powered. The majority included more than 100 patients (6/10). Four were prognostic (focused on either the prediction of bleeding volume or the prediction of needed management) while six diagnostic (PAS vs not PAS classification) studies. The median RQS was 8, with maximum and minimum respectively equal to 17/36 and - 6/36. Major methodological concerns were the lack of feature stability to multiple segmentation testing and poor data openness.

CONCLUSIONS

Radiomics studies focused on PAS disorders showed a heterogeneous methodological quality, overall lower than desirable. Furthermore, many relevant research questions remain unexplored. More robust investigations are needed to foster advancements in the field and possibly clinical translation.

A computerized diagnostic model for automatically evaluating placenta accrete spectrum disorders based on the combined MR radiomics-clinical signatures

We aimed to establish a computerized diagnostic model to predict placenta accrete spectrum (PAS) disorders based on T2-weighted MR imaging. We recruited pregnant women with clinically suspected PAS disorders between January 2015 and December 2018 in our institution. All preoperative T2-weighted imaging (T2WI) MR images were manually outlined on the picture archive communication system terminal server. A nnU-Net network for automatic segmentation and the corresponding radiomics features extracted from the segmented region were applied to build a radiomics-clinical model for PAS disorders identification. Taking the surgical or pathological findings as the reference standard, we compared this computerized model’s diagnostic performance in detecting PAS disorders. In the training cohort, our model combining both radiomics and clinical characteristics yielded an accuracy of 0.771, a sensitivity of 0.854, and a specificity of 0.750 in identifying PAS disorders. In the testing cohort, this model achieved a segmentation mean Dice coefficient of 0.890 and yielded an accuracy of 0.825, a sensitivity of 0.830 and a specificity of 0.822. In the external validation cohort, this computer-aided diagnostic model yielded an accuracy of 0.690, a sensitivity of 0.929 and a specificity of 0.467 in identifying placenta increta. In the present study, a machine learning model based on preoperative T2WI-based imaging had high accuracy in identifying PAS disorders in respect of surgical and histological findings.

Qualitative evaluation of MR images for assessing placenta accreta spectrum disorders in patients with placenta previa: A pilot validation study

PURPOSE

To validate a qualitative imaging method using magnetic resonance (MR) for predicting placental accreta spectrum (PAS) in patients with placenta previa (PP).

METHOD

Two MR imaging methods built in our previous experience was tested in an external comparable group of sixty-five patients with PP; these methods consisted of presence of at least one (Method 1) or two (Method 2) of the following abnormal MR imaging signs: intraplacental dark bands, focal interruption of myometrial border and abnormal placental vascularity. Three groups of radiologists with different level of expertise evaluated MR images: at least 5 years of experience in body imaging (Group 1); at least 10 (Group 2) or 20 (Group 3) years of experience in genito-urinary MR. While radiologists of Group 1 routinely evaluated MR images, those of Groups 2 and 3 used both Methods 1 and 2.

RESULTS

A significant (p < 0.005) difference was found between the diagnostic accuracy values of imaging evaluation performed by Group 3 using Method 1 (63%) and Method 2 (89%); of note, the accuracy of Method 2 by Group 3 was also significantly (p < 0.005) higher compared to that of both Methods 1 (46%) and 2 (63%) by Group 2 as well as to that of the routine evaluation by Group 1 (60%).

CONCLUSIONS

The qualitative identification of at least two abnormal MR signs (Method 2) represents an accurate method for predicting PAS in patients with PP particularly when this method was used by more experienced radiologists; thus, imaging expertise and methodology is required for this purpose.

Development and validation of a magnetic resonance imaging-based nomogram for predicting invasive forms of placental accreta spectrum disorders

AIM

The aim of the study was to develop and validate a magnetic resonance imaging (MRI)-based nomogram for predicting invasive forms of placental accreta spectrum (PAS) disorders (placenta increta and percreta) with "uncertain ultrasound diagnosis."

METHODS

This was a retrospective cohort study of a primary cohort of 118 patients and a validation cohort of 65 patients with "uncertain ultrasound diagnosis," who were further evaluated by MRI. MRI signs associated with PAS disorders were analyzed between invasive and noninvasive groups by both univariate and logistic regression to construct the nomogram. The accuracy and discriminative ability of the nomogram were measured by concordance index (C-index) and calibration curve internally and externally.

RESULTS

The history of previous cesarean deliveries (odds ratio [OR], 3.27; 95% confidence interval [CI], 1.16-9.27), loss of double-line sign (OR, 9.49; 95% CI, 3.06-29.48), abnormal uterine bulging (OR, 4.05; 95% CI, 1.53-10.69), and disorganized abnormal placenta vascularity (OR, 3.38; 95% CI, 1.09-10.50) were imputed for the nomogram. The C-index of the nomogram was 0.85 for internal validation and 0.84 for external validation. Calibration curve showed good agreement with predicted risk and actual observation for both primary and validation cohort.

CONCLUSIONS

MRI can be a useful adjunct for clinical staging of patients with "uncertain ultrasound diagnosis."

Prediction of placenta accreta spectrum in patients with placenta previa using clinical risk factors, ultrasound and magnetic resonance imaging findings

OBJECTIVES

To predict placental accreta spectrum (PAS) in patients with placenta previa (PP) evaluating clinical risk factors (CRF), ultrasound (US) and magnetic resonance imaging (MRI) findings.

METHODS

Seventy patients with PP were retrospectively selected. CRF were retrieved from medical records. US and MRI images were evaluated to detect imaging signs suggestive of PAS. Univariable analysis was performed to identify CRF, US and MRI signs associated with PAS considering histology as standard of reference. Receiver operating characteristic curve (ROC) analysis was performed, and the area under the curve (AUC) was calculated. Multivariable analysis was also performed.

RESULTS

At univariable analysis, the number of previous cesarean section, smoking, loss of the retroplacental clear space, myometrial thinning < 1 mm, placental lacunae, intraplacental dark bands (IDB), focal interruption of myometrial border (FIMB) and abnormal vascularity were statistically significant. The AUC in predicting PAS progressively increased using CRF, US and MRI signs (0.69, 0.79 and 0.94, respectively; p < 0.05); the accuracy of MRI alone was similar to that obtained combining CRF, US and MRI variables (AUC = 0.97) and was significantly higher (p < 0.05) than that combining CRF and US (AUC = 0.83). Multivariable analysis showed that only IDB (p = 0.012) and FIMB (p = 0.029) were independently associated with PAS.

CONCLUSIONS

MRI is the best modality to predict PAS in patients with PP independently from CRF and/or US finding. It is reasonable to propose the combined assessment of CRF and US as the first diagnostic level to predict PAS, sparing MRI for selected cases in which US findings are uncertain for PAS.

In-vitro fertilisation-embryo-transfer complicates the antenatal diagnosis of placenta accreta spectrum using MRI: a retrospective analysis

AIM

To evaluate the diagnostic accuracy of magnetic resonance imaging (MRI) for the antenatal diagnosis of placenta accreta spectrum (PAS).

MATERIALS AND METHODS

Data from 95 patients with placenta previa or low-lying placenta who underwent MRI at Osaka University Hospital for the antenatal diagnosis of PAS between January 2013 and December 2018 were reviewed retrospectively. The antenatal MRI signs suggesting PAS were assessed. Patients were divided into two groups depending on whether they were diagnosed with PAS. Factors that affected PAS diagnosis were identified using multivariate analysis.

RESULTS

The diagnostic accuracy of MRI for detecting PAS was as follows: 71.4% sensitivity, 96.4% specificity, and area under the curve (AUC) of 0.839 (95% confidence interval [CI]: 0.73-0.91). The diagnostic accuracy was lower in patients with in-vitro fertilisation with embryo transfer (IVF-ET): 22.2% sensitivity, 93.3% specificity, and AUC=0.578 (95% CI: 0.417-0.724). On multivariate analysis, only IVF-ET showed a significant association with false-positive or -negative MRI diagnosis of PAS (adjusted odds ratio: 26.5; 95% CI: 2.42-289.4; p=0.007).

CONCLUSION

IVF-ET affects the antenatal diagnosis of PAS using MRI.

Correlation of Placental Magnetic Resonance Imaging With Histopathologic Diagnosis: Detection of Aberrations in Structure and Water Diffusivity

OBJECTIVE

Noninvasive methods to identify placental pathologic conditions are being sought in order to recognize these conditions at an earlier stage leading to improved clinical interventions and perinatal outcomes. The objective of this study was to examine fixed tissue slices of placenta by T₂- and diffusion-weighted magnetic resonance imaging (MRI) and correlate the images with placental pathologic findings defined by routine gross and histologic examination.

METHODS

Four formalin-fixed placentas with significant placental pathology (maternal vascular malperfusion, chronic villitis of unknown etiology, and massive perivillous fibrin deposition) and 2 histologically normal placentas were evaluated by high-resolution MRI. Representative placental slices were selected (2 cm long and 10 mm wide) and rehydrated. Imaging was performed on a Bruker Avance 14.1 T microimager. Diffusion-weighted images were acquired from 16 slices using slice thickness 0.5 mm and in-plane resolution approximately 100 µm × 100 µm. T₂ maps were obtained from the same slices. T₂ relaxation time and apparent diffusion coefficient (ADC) were acquired from representative regions of interest and compared between normal and diseased placentas.

RESULTS

In T₂- and diffusion-weighted images, the placental microstructure differed subjectively between diseased and normal placentas. Furthermore, diseased placentas showed statistically significantly longer mean T₂ relaxation times and generally higher mean ADC.

CONCLUSION

Diffusion- and T₂-weighted MRI can potentially be used to detect significant placental pathology by using T₂ relaxation time and ADC as markers of altered placental microstructure.

MRI of placenta accreta: diagnostic accuracy and impact of interventional radiology on foetal-maternal delivery outcomes in high-risk women

OBJECTIVE

To assess accuracy and reproducibility of MRI diagnosis of invasive placentation (IP) in high-risk patients and to evaluate reliability of MRI features. Secondary aim was to evaluate impact of interventional radiology (IR) on delivery outcomes in patients with IP at MRI.

METHODS

26 patients (mean age 36.24 y/o,SD 6.16) with clinical risk-factors and echographic suspicion of IP underwent 1.5 T-MRI. Two readers reviewed images. Gold-standard was histology in hysterectomised patients and obstetric evaluation at delivery for patients with preserved uterus. Accuracy and reproducibility of MRI findings were calculated.

RESULTS

Incidence of IP was 50% (13/26) and of PP was 11.54% (3/26). MRI showed 100% sensitivity (95% CI = 75.3-100%) and 92.3% specificity (95% CI = 64.0-100%) in the diagnosis of IP. Gold-standard was histology in 10 cases and obstetric evaluation in 16. MRI findings with higher sensitivity were placental heterogeneity, uterine bulging and black intraplacental bands. Uterine scarring, placental heterogeneity, myometrial interruption and tenting of the bladder showed better specificity. MRI inter-rater agreement with Cohen's K was 1. 11 patients among 14 with MRI diagnosis of IP received IR assistance with positive impact on delivery outcomes in terms of blood loss, red cells count, intense care unit length of stay, days of hospitalisation and risk of being transfused.

CONCLUSION

MRI is an accurate and reproducible technique in prenatal diagnosis of IP. MRI helps planning a safe and appropriate delivery eventually assisted by IR, which positively affects foetal and maternal outcomes.

ADVANCES IN KNOWLEDGE

The adoption of MRI evaluation in patients with high risk of invasive placentation allows a more accurate diagnosis in terms of both presence of the disease and its extension to or through or even beyond the myometrium. This led to a better dedicated delivery management with eventual adoption of interventional radiology with a global positive effect on foetal and maternal outcomes.

US and MR imaging findings to detect placental adhesion spectrum (PAS) in patients with placenta previa: a comparative systematic study

PURPOSE

To compare the performance US and MR in identifying placental adhesion spectrum (PAS) in placenta previa (PP) and to establish a potential method of image interpretation.

METHODS

US and MR examinations of 51 patients with PP were selected. The presence of imaging signs commonly used to detect PAS was assessed. Penalized logistic regression was performed considering histology as standard of reference; only signs statistically significant (p < 0.05) were considered for ROC and multivariate analysis. The probability of PAS according to the presence of US and/or MR signs was then assessed.

RESULTS

At univariate analysis, loss of retroplacental clear space, myometrial thinning (MT) and placenta lacunar spaces on US, intraplacental dark bands (IDBs), focal interruption of myometrial border (FIMB) and abnormal vascularity (AV) on MR were statistically significant (p < 0.01). Three diagnostic methods for PAS were then developed for both US and MR when at least one (Method 1), two (Method 2) or three (Method 3) imaging signs occurred, respectively. Method 2 for MR showed a significantly (p < 0.05) higher accuracy (91%) compared to the other methods. When MR IDBs and AV as well as IDBs and FIMB were present in combination with US MT the probability of PAS increased from 75 to 90% and from 80 to 91%, respectively.

CONCLUSION

MR demonstrated a higher diagnostic accuracy than US to detect PAS. However, since the combination of MR and US signs could improve the probability to detect PAS, a complementary diagnostic role of these techniques could be considered.

Machine learning analysis of MRI-derived texture features to predict placenta accreta spectrum in patients with placenta previa

PURPOSE

To evaluate whether a machine learning (ML) analysis employing MRI-derived texture analysis (TA) features could be useful in assessing the presence of placenta accreta spectrum (PAS) in patients with placenta previa (PP). The hypothesis is that TA features may reflect histological abnormalities underlying PAS in patients with PP thus helping in differentiating positive from negative cases.

MATERIALS AND METHODS

Pre-operative MRI examinations of 64 patients with PP of which 20 positive (12 accreta, 7 increta and 1 percreta) and 44 negative for PAS were retrospectively selected. Multiple (n = 3) rounded regions of interest (ROIs) were manually positioned on sagittal or coronal T2-weighted images over homogeneous placental tissue close to the placental-myometrial interface for each patient to extract TA features. After balancing the dataset with the Synthetic Minority Over-sampling Technique, training and testing sets were obtained using Hold-out with a 75/25% split. Different algorithms were applied on the training set using the wrapper method, which looks for the best combination of features based on the optimization of a heuristic function in order to get the highest accuracy, and a 10-fold Cross-validation. The accuracy of the best models was also assessed on the test set. Histology was used as the standard of reference.

RESULTS

A total of 192 ROIs were positioned and a ROI-based analysis was then conducted. Among the different algorithms, k-nearest neighbors obtained the highest accuracy (98.1%), precision (98.7%), sensitivity (97.5%) and specificity (98.7%) while exploiting the lowest number of features (n = 26); conversely, the Naïve Bayes algorithm got the lowest scores showing an accuracy of 80.5%.

CONCLUSION

ML analysis using MRI-derived TA features could be a feasible tool in the identification of placental tissue abnormalities underlying PAS in patients with PP. This approach might represent an additional tool in the clinical practice, thus expanding the application field of artificial intelligence to medical images.