Predicting Cleft Lip in Unborn Babies Using Ultrasound Images and Machine Learning

Abstract: Predicting cleft lip in unborn babies using ultrasound images and machine learning offers a non- invasive approach for early detection during prenatal stages. The system uses deep learning to automatically identify facial irregularities in ultrasound scans, focusing on the lip and surrounding regions. A CNN model trained on annotated images classifies scans as normal or cleft-affected with high accuracy, supported by preprocessing steps such as noise reduction and contrast enhancement. Grad-CAM visualizations highlight the regions influencing predictions, improving clinical transparency.

A Python-based web application built using Flask or Django allows users to upload ultrasound images and receive real- time predictions with confidence scores. This cost-effective tool supports gynecologists and radiologists in early anomaly detection, reducing dependence on postnatal procedures. Overall, the system strengthens prenatal screening and contributes to timely intervention and improved neonatal outcomes.

Keywords: Cleft lip detection, ultrasound img, deep learning, CNN, Grad-CAM, medical imaging, AI healthcare.