Radiomics-based prediction of FIGO grade for placenta accreta spectrum

Area of Science:

• Radiology
• Oncology
• Medical Imaging

Background:

• Placenta accreta spectrum (PAS) is a rare but life-threatening pregnancy complication.
• Accurate prediction of PAS severity is crucial for individualized birth planning.
• Distinguishing between invasive and non-invasive PAS subtypes antenatally remains a clinical challenge.

Purpose of the Study:

• To investigate the potential of radiomic analysis of T2-weighted MRI for predicting severe PAS.
• To differentiate between histopathological subtypes of PAS using imaging features.
• To develop a predictive model for severe (FIGO grade 3) PAS.

Main Methods:

• Retrospective analysis of a prospective cohort (2018-2022) of women with histologically confirmed PAS who underwent antenatal MRI.
• Extraction of radiomic features from T2-weighted MRI sequences.
• Application of univariate and multivariate regression analyses, including a support vector machine model, to predict severe PAS (FIGO grade 3).

Main Results:

• Univariate analysis showed a sensitivity of 0.64, specificity of 0.93, accuracy of 0.58, and AUC of 0.77 for predicting severe PAS.
• A multivariate support vector machine model achieved a sensitivity of 0.30, specificity of 0.74, accuracy of 0.58, and AUC of 0.53.
• Forty-one women met the inclusion criteria for the study.

Conclusions:

• Radiomic analysis of T2-weighted MRI demonstrates potential for predicting severe PAS cases.
• The developed machine learning pipeline shows promise in classifying severe PAS subtypes antenatally.
• This imaging-based approach can contribute to safer and more individualized care planning for pregnancies affected by PAS.