Enhancing Spina Bifida Detection in Fetal Ultrasound Images using Generative Adversarial Networks

Spina bifida is a congenital neurological condition caused by the incomplete closure of the spinal cord during fetal development, potentially leading to severe complications such as permanent paralysis if untreated. Early and accurate detection is critical for timely intervention, with ultrasonography serving as the gold standard for fetal monitoring. This research employs a Conditional Generative Adversarial Network (cGAN) to detect spina bifida by utilizing clinical features such as gestational age, maternal age, spinal lesion size, and head circumference. The cGAN architecture comprises a generator that synthesizes realistic feature sets conditioned on spina bifida presence and a discriminator that classifies the condition while distinguishing between real and synthetic data. Adversarial training, combined with careful normalization and preprocessing of the input data, ensures the generation of clinically relevant synthetic datasets that enhance diversity and robustness. This approach augments limited real-world data, facilitating robust classification and underscoring the potential of cGANs in advancing data-driven prenatal diagnostics and improving early spina bifida detection.