Background: Fetal growth restriction (FGR) and placental insufficiency (PI) are major contributors to adverse perinatal outcomes, and early identification remains clinically challenging. This study aimed to evaluate the utility of artificial intelligence (AI)-based fetal magnetic resonance imaging (MRI) for early screening of FGR and PI. Methods: A retrospective analysis was conducted on 120 pregnant women with a gestational age of 24–36 weeks who underwent fetal MRI examinations at the Second Hospital of West China, Sichuan University, from September 2024 to May 2025. Participants were divided into an FGR group, defined by an estimated fetal weight (EFW) below the 10th percentile for gestational age based on the Hadlock growth chart (n = 60), and a non-FGR group, defined by an EFW at or above the 10th percentile for gestational age (n = 60). MRI examinations were performed using a 3.0-T MRI Premier scanner with a 32-channel Head GE AIR™ Coil body coil, acquiring T2-weighted single-shot fast spin-echo (SSFSE) sequences of the fetal abdomen and placenta. Accelerated protocol images were reconstructed using the AIR™ Recon Deep Learning (DL) algorithm. FGR assessment relied on EFW, calculated from U-Net-segmented fetal abdominal volume combined with ultrasound-measured fetal femur length (FL) using the Hadlock formula, and compared with birth weight. EFW below the 10th percentile for gestational age served as the criterion for FGR. Segmentation performance was evaluated using the Dice similarity coefficient (target >0.85), diagnostic accuracy via the area under the receiver operating characteristic (ROC) curve (AUROC), and EFW error via mean absolute error (MAE, target <150 g). Ultrasound FL measurements were performed using a GE Voluson E10 system by certified sonographers, following the International Society of Ultrasound in Obstetrics and Gynecology (ISUOG) guidelines, within 3 days of MRI to ensure consistency. Results: AIR™ Recon DL-reconstructed images achieved a signal-to-noise ratio (SNR) of 59.5 ± 7.1 and a contrast-to-noise ratio (CNR) of 26.9 ± 4.8, significantly higher than non-reconstructed images (p < 0.05). The Dice coefficient for placental and fetal structure segmentation was 0.892 ± 0.021, the EFW MAE was 148.2 g, and FGR diagnostic accuracy was 91.7% (AUROC = 0.938), outperforming manual assessment (84.5%, AUROC = 0.864; p = 0.032). In the FGR with PI subgroup, diagnostic accuracy reached 93.4%, with an MAE of 135.6 g (p = 0.016). Conclusions: Fetal MRI combined with the AIR™ Recon DL algorithm shows significant clinical value in assessing FGR and PI. This system enhances image quality, accelerates scanning, and reduces artifacts, thus improving the accuracy of FGR and PI diagnosis in early screening. This technique also offers reliable support for early screening and the development of individualized treatment plans in the context of high-risk pregnancy.
Zhang et al. (Mon,) studied this question.