Detecting Fetal Brain Abnormalities Using Deep Learning Technique

In this study, the critical issue of fetal brain abnormalities which can cause serious health problems and long-term challenges for affected individuals is addressed. Early detection and treatment are essential for a better prognosis and hence this work has suggested a deep learning-based method for detecting fetal brain abnormalities from ultrasound images. Leveraging the power of convolutional neural networks (CNNs), the selected pre-trained CNN models are extended with 100 additional layers and fine-tuned to enhance their performance for fetal brain anomaly classification. Transfer learning model such as MobileNetV2 with 88 layers, VGG16 with 16 layers, ResNet50 with 50 layers and InceptionV3 with 159 layers are trained using a combination of two publicly accessible datasets comprising ultrasound images of fetal brains. The results demonstrate that our approach, particularly employing the MobileNetV2 architecture, achieves a remarkable classification accuracy of 90% for fetal brain disorders, surpassing VGG16, ResNet50, and InceptionV3.This encouraging result highlights the utilization of deep learning in early identification and treatment, improving the long-term prognosis of infants with fetal brain abnormalities and thus enhancing prenatal care.