FB-CBF Net: Fetal Brain Maturity Based Gestational Age Estimation using Cuckoo Catfish Optimized Bidirectional Feature Pyramid Network

Accurate estimation of fetal gestational age (GA) is essential for monitoring neurodevelopment and identifying abnormalities during pregnancy. However, ultrasound-based biometry becomes less reliable in later gestation and fetal growth restriction cases by limiting accurate brain maturity assessment. To overcome this issue, a novel FB-CBF Net has been proposed for fetal gestational age estimation from MRI based on the cerebral cortical sulcation. The Retinex Bilateral (REB) filter enhances MRI image quality by improving contrast and preserving anatomical structures. The Swin U-Net is employed to segment fetal brain regions and preserve fine cortical patterns associated with developmental progression. A bidirectional attention-based feature pyramid network (BHA-FPN) extracts cortical biomarkers, including features for sulcal depth, cortical thickness, and gray-to-white matter intensity. The Cuckoo Catfish Optimization (CCO) algorithm is employed to select the most discriminative features by reducing feature redundancy. Finally, the optimized feature set is fed into a fully connected layer (FCL) to classify gestational age stages such as Parieto-Occipital Sulcus (POS), Calcarine Sulcus (CS), Central Sulcus (CES), Superior Temporal Sulcus (STS), and Pre-and Post-Central Sulcus (PPCS). The BHA-FPN was developed using T2-weighted MRI images of 659 fetal brains. Predicting fetal brain age was achieved with a mean absolute error (MAE) of 0.95 and an accuracy of 99.05%. Moreover, the proposed FB-CBF Net increases the overall accuracy by 11.29%, 4.98%, 3.97%, and 1.24% compared to MTSE U-Net, BAPNet, Attention-guided CNN, and ResNet-50, respectively.