Early Detection of Fetal Skull Anomalies using Web-Based Convolutional Neural Networks

Congenital fetal anomalies occur between 2-3 % of pregnancies globally. Approximately 20 % of anomalies have a genetic issue, including those caused by underlying chromosomal abnormalites or an individual gene defect. Mainly there are 2 types of fetal anomalies structural anomalies and functional anomalies, structural anomalies influence the growing baby’s heart, kidneys, lungs, limbs and facial features and functional anomalies influence the working of an organ or system, such as the brain.This research proposes an approach to improve quality of prenatal diagnosis for early detection of fetal anomalies. This paper proposed a convolutional neural network-based deep learning technique to identify Cloverleaf anomaly in the developing skull using information from ultrasound images and the Transfer Learning method. Ultrasound data is used to compare between different architectures VGG16, ResNet50, MobileNet-v2 and Proposed Algorithm to find the optimal CNN. Experimental results based on Transfer Learning indicated that the Proposed Algorithm outperformed the pre-trained CNN models (MobileNet-v2, ResNet50, and VGG16). Additionally, created the Flask web application, which provides a unified platform for detecting prenatal anomalies. The app, with its user-friendly design and prediction capabilities, is a valuable resource for medical experts and expectant parents looking for accurate prenatal diagnoses. The proposed model might be useful for early detection of prenatal skull anomalies.