Radiomic study of antenatal prediction of severe placenta accreta spectrum from MRI

Abstract

Objectives: We previously demonstrated the potential of radiomics for the prediction of severe histological placenta accreta spectrum (PAS) subtypes using T2-weighted MRI. We aim to validate our model using an additional dataset. Secondly, we explore whether the performance is improved using a new approach to develop a new multivariate radiomics model.

Methods: Multi-centre retrospective analysis was conducted between 2018 and 2023. Inclusion criteria: MRI performed for suspicion of PAS from ultrasound, clinical findings of PAS at laparotomy and/or histopathological confirmation. Radiomic features were extracted from T2-weighted MRI. The previous multivariate model was validated. Secondly, a 5-radiomic feature random forest classifier was selected from a randomized feature selection scheme to predict invasive placenta increta PAS cases. Prediction performance was assessed based on several metrics including area under the curve (AUC) of the receiver operating characteristic curve (ROC), sensitivity, and specificity.

Results: We present 100 women [mean age 34.6 (±3.9) with PAS], 64 of whom had placenta increta. Firstly, we validated the previous multivariate model and found that a support vector machine classifier had a sensitivity of 0.620 (95% CI: 0.068; 1.0), specificity of 0.619 (95% CI: 0.059; 1.0), an AUC of 0.671 (95% CI: 0.440; 0.922), and accuracy of 0.602 (95% CI: 0.353; 0.817) for predicting placenta increta. From the new multivariate model, the best 5-feature subset was selected via the random subset feature selection scheme comprised of 4 radiomic features and 1 clinical variable (number of previous caesareans). This clinical-radiomic model achieved an AUC of 0.713 (95% CI: 0.551; 0.854), accuracy of 0.695 (95% CI 0.563; 0.793), sensitivity of 0.843 (95% CI 0.682; 0.990), and specificity of 0.447 (95% CI 0.167; 0.667).

Conclusion: We validated our previous model and present a new multivariate radiomic model for the prediction of severe placenta increta from a well-defined, cohort of PAS cases.

Advances in knowledge: Radiomic features demonstrate good predictive potential for identifying placenta increta. This suggests radiomics may be a useful adjunct to clinicians caring for women with this high-risk pregnancy condition.

Keywords: MRI; machine learning; placenta accreta spectrum; pregnancy; radiomics.

Figure 1.

Histopathology of accreta and increta case and MRI segmentations. (A) Haematoxylin and eosin stained slide at high 200 µm power. Placental villi are seen in direct contact with the myometrium with no decidua present in a case of histopathological placenta accreta. (B) Haematoxylin and eosin stained slide at high 200 µm power. Placental villi are seen deep within the myometrium with no intervening decidua in a case of histopathological placenta increta. (C and D) An example of the placental regions of interest from which radiomic features were extracted. (C) Sagittal, b-SSFP sequence, of PAS case at 30 weeks’ gestation. The placenta is anterior completely covering the internal cervical os (placenta previa). (D) Inferior (lower, purple) and superior (upper, green) placental regions of interest. PAS = placenta accreta spectrum.

Figure 2.

Steps in radiomic processing. A summary of the radiomic workflow.

Figure 3.

(A) Heatmap showing association of PAS grade and radiomic features. (B) Scaled PCA of radiomic feature set, assessing distribution of PAS grade cases with respect to the first 2 principle components, “Y” indicates placenta increta cases. (C) Scaled PCA of radiomic features by MRI scanner. (D and E) The overall results of the univariate analysis for the inferior and superior placental region, the boxplots in (D) show the performance metrics of the best-performing radiomic features and (E) summarizes the results from the univariate analysis, including the percentage of radiomic features with a performance metric greater than 65%. PAS = placenta accreta spectrum, PCA = principle component analysis, AUC = area under the curve. Sens = sensitivity, Spec = specificity, Acc = accuracy.

Figure 4.

(A and B) The performance of the 5-feature random forest classifier selected from the random subset feature selection scheme. The figures show a comparison of these performance metrics with those obtained from a traditional logistic regression model, to illustrate the benefit of using a typical machine learning modelling approach in order to leverage the potential of radiomic features effectively. (A) ROC curves of the fitted random forest (blue) and logistic regression (red) models for placenta increta classification. (B) Distributions of cross-validated performance metrics of the 5-feature random forest placenta increta classifier obtained via random subset feature selection (in blue), and the same distributions obtained by using a conventional logistic regression model with the same features (in red). (C) The calibration curve for the multivariate model for predicted and observed placenta increta cases. The curve indicates there was good agreement between model prediction and observed rates for both placenta increta and placenta accreta cases. PAS = placenta accreta spectrum, GLM = generalized linear model, RF = random forest.