A LASSO-Based Ultrasound–MRI Scoring Model for Predicting Severe Maternal Outcomes in Placenta Accreta Spectrum

Objective: To evaluate the predictive value of a weighted scoring model based on LASSO logistic regression integrating ultrasound and magnetic resonance imaging (MRI) for severe adverse maternal outcomes (SAMO) in patients with placenta accreta spectrum (PAS). SAMO was defined as hysterectomy and/or postpartum hemorrhage ≥1500 mL within 24 hours. Methods: A total of 133 patients with PAS who delivered at our hospital between January 2020 and December 2022 were retrospectively enrolled. Patients were divided into the SAMO group (n=49) and the non-SAMO group (n=84). Twenty-one prenatal ultrasound and MRI imaging features were evaluated. LASSO logistic regression was used to identify independent predictors and to construct a weighted scoring model. Model performance was assessed using the receiver operating characteristic (ROC) curve. Clinical utility was evaluated using decision curve analysis (DCA) to quantify net benefit across different risk thresholds. Results: LASSO logistic regression initially selected 11 imaging features with non-zero coefficients from the 21 candidate variables. After excluding three features with negligible coefficients (assigned zero weight in the simplified scoring system), eight predictors were retained to construct an 11-point weighted scoring model: anterior placental location, placenta previa, placental thickness ≥5 cm, placental lacunae, cervical length ≤2.5 cm, lower uterine segment myometrial thickness ≤1 mm, intraplacental T2-weighted (T2WI) low-signal bands, and the “tenting sign.” These predictors were assigned weights of 1–2 points to construct an 11-point scoring system. The model achieved an AUC of 0.879 (95% CI: 0.812–0.945) with a sensitivity of 75.51% and a specificity of 88.10% at the optimal cutoff of 7 points. Patients with a score ≥7 had a significantly higher risk of SAMO (OR=22.82, 95% CI: 9.03–57.68). DCA demonstrated a positive net clinical benefit across a wide range of threshold probabilities (approximately 0.05–0.85), substantially outperforming both the “treat-all” and “treat-none” strategies. Conclusion: This internally validated ultrasound–MRI combined scoring system provides a clinically practical tool for early risk stratification of severe maternal outcomes in patients with PAS. The model demonstrates robust predictive performance and clear clinical utility. Future external validation studies are warranted to confirm its generalizability across diverse healthcare settings.